Download Gasterosteus aculeatus L.

Blackwell Science, LtdOxford, UKBIJBiological Journal of the Linnean Society0024-4066The Linnean Society of London, 2004? 2004
822
189203
Original Article
DIVERGENCE OF STICKLEBACKS
L. I. DOUCETTE
Et al.
Biological Journal of the Linnean Society, 2004, 82, 189–203. With 6 figures
Risk of predation as a promoting factor of
species divergence in threespine sticklebacks
(Gasterosteus aculeatus L.)
LISA I. DOUCETTE1,2*, SKÚLI SKÚLASON1 and SIGURDUR S. SNORRASON2
1
Hólar College, Hólar, Hjaltadal, 551 Sauð arkrokur, Iceland
Institute of Biology, University of Iceland, Grensásvegur 11, 108 Reykjavik, Iceland
2
Received 7 April 2003; accepted for publication 17 December 2003
Icelandic freshwater systems are geologically young and contain only six species of freshwater fish. As these species
colonized Icelandic fresh waters they were presented with a diversity of unique, uncontested habitats and food
resources, promoting the evolution of new behaviour strategies crucial to the formation of new morphs and speciation. To determine the likelihood that predation threat could affect the antipredator behaviour and possibly the sympatric divergence of prey populations, we analysed antipredator behaviour of seven groups of Icelandic threespine
sticklebacks (Gasterosteus aculeatus): two marine groups, one group from a lake without piscine predators, and two
polymorphic lake populations, each with two groups occupying unique habitats. Shoaling cohesion, school formation
and duration, and vigilance in predator inspection/avoidance behaviour varied greatly among groups. The differences
appeared to be related to the risk of predation as well as to opportunities and constraints set by the different habitats. Antipredator behaviour was especially pronounced and differed extensively in two polymorphic forms from the
lake Thingvallavatn, where predation risk is very high. By keeping the two morphs separate in their respective habitats, high predation risk may be a contributing factor in promoting the habitat-specific divergence of G. aculeatus
seen in the lake. This suggests that in situations where refuge habitats are spatially separated, the risk of predation
may contribute to the evolution of separate sympatric forms of small fish such as G. aculeatus. © 2004 The Linnean
Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203.
ADDITIONAL KEYWORDS: adaptation – antipredator behaviour – apparent competition – inspection –
schooling – shoaling – speciation – startle response – sympatric.
INTRODUCTION
The reduction of gene flow and isolation between subpopulations as they specialize for a novel resource
such as an alternative habitat (Maynard Smith, 1966;
Rosenzweig, 1987), food source (Skúlason et al., 1993;
Smith, 1993), microclimate (Koopman, 1949), or host
species (Bush, 1975) will lead to speciation without
the need for geographic isolation (Dobhansky, 1940;
Bush, 1994). Many models of sympatric speciation are
based on the idea that speciation can occur when individuals of a population have ready access to alternative habitats within a region (Maynard Smith, 1966;
*Corresponding author. Current address: Zoology, University
of New England, Armidale NSW 2351, Australia. E-mail:
[email protected]
Johnson & Gullberg, 1998). These models generally
assume that each individual in the population selects
one alternative habitat and then lives and utilizes
resources in that habitat almost exclusively (Johnson
& Gullberg, 1998). This can create a biological barrier
to interbreeding (Futuyma, 1986) and result in selection against intermediate phenotypes. Selection can
be disruptive, acting uniformly on every individual, or
divergent in two or more habitats, allowing the
extreme phenotypes to exploit the habitats where they
best fit (Kondrashov, Yampolsky & Shabalina, 1998).
Apart from alternative habitats, other physical and
biotic factors can affect the likelihood of sympatric
morph divergence and speciation, such as stability of
alternative habitats, productivity, distribution of
potential breeding sites and predation risk (Skúlason,
Snorrason & Jónsson, 1999; Snorrason & Skúlason,
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
189
190
L. I. DOUCETTE ET AL.
2004). Whether the risk of predation is a selective
force that accelerates or retards the process of speciation is still a matter of controversy (McPhail, 1969;
Holt, 1977; Schluter, 1998). Some researchers have
suggested that the absence of competitors and predators in novel environments may lead to high population densities of colonizers, and that this ‘ecological
persistence’ allows them to avoid extinction long
enough to evolve reproductive isolation (Mayr, 1963;
Schluter, 1998). Others have argued that predation
relaxes competitive pressures by reducing the density
of competing prey species, allowing higher diversity of
prey species than in predator-free environments (Holt,
1977). Holt (1977) modelled different levels of competition and predation in prey communities and concluded that in most natural communities patterns of
abundance probably reflect both direct competition
and ‘apparent competition’, under which prey compete
for predator-free space. However, in natural communities in which direct competition is weak or entirely
absent, apparent competition may play a primary role
in the structure of the community. Other researchers
agree that apparent competition may be a strong
diversifying force (Brown & Vincent, 1992; Abrams,
2000). Doebeli & Dieckmann (2000) showed through
modelling that apparent competition could lead to
sympatric speciation through evolutionary branching
as prey species developed specialized defence morphologies and behaviour. Predation has already been
shown to be a strong selective force for male breeding
coloration (McPhail, 1969; Endler, 1995), antipredator
morphology (Walker, 1997), and feeding (Peeke & Morgan, 2000), schooling (Seghers, 1974; Magurran,
1999), and inspection behaviour (Magurran & Seghers, 1994) in several fish species.
Due to geographical isolation and the limited
age of the freshwater systems (maximum
age = 11 000 years), Iceland harbours only six species
of freshwater fish, which often experience high
intraspecific but low interspecific competition. Lakes
in the neovolcanic zone often offer distinct benthic
habitats, such as lava formations with deep rifts,
structurally complex lava rubble with extensive subbenthic spaces that smaller fish can exploit, and areas
of spring water inflow at constant, low temperatures.
These novel conditions are thought to have promoted
resource specialization and evolution of benthic morphs of Arctic charr Salvelinus alpinus (Skúlason et al.,
1992, 1999; Snorrason et al., 1994). Recent studies in
Iceland, have revealed significant differences in morphology and foraging behaviour between threespine
sticklebacks G. aculeatus in discrete habitats in some
volcanic lakes, a situation that seems to parallel the
morph formation in Salv. alpinus (Doucette, 2001;
Kristjánsson, 2001; Kristjánsson, Skúlason & Noakes,
2002a).
We examined shoaling density and antipredator
behaviour of G. aculeatus from seven different sites in
south-western Iceland to determine if sympatric and
allopatric groups employed different antipredator tactics. Two volcanic lakes, Thingvallavatn and Frostastaðavatn (the suffix ‘vatn’ means lake in Icelandic),
having separate lava littoral and soft-bottom, vegetated habitats, were selected for studying sympatric
divergence. G. aculeatus from these two habitats in
Thingvallavatn have been found to differ in morphology and foraging behaviour, but less clear morphological and behavioural differences were observed in
Frostastaðavatn (Doucette, 2001; Kristjánsson, 2001;
Kristjánsson et al., 2002a). Based on diet analyses of
Salv. alpinus, predation pressure on G. aculeatus is
considerable in Thingvallavatn (Malmquist et al.,
1992), but is less intense in Frostastaðavatn (B. Jónsson, pers. comm.). By comparing the antipredator
behaviour of the sympatric groups within these two
lakes, with that of two ancestral marine groups, and a
group from a third lake containing no predators, we
searched for selection signatures of predation risk and
assessed its role in the process of morph formation and
sympatric speciation in G. aculeatus.
MATERIAL AND METHODS
G. ACULEATUS
COLLECTION AND MAINTENANCE
During the summer of 1999, Gasterosteus aculeatus
were collected from seven sites throughout Iceland
(Table 1). Thingvallavatn is a very large (83 km2),
deep (mean depth = 34 m) lake in the south of Iceland
(64∞10¢N, 21∞10¢W) (Adalsteinsson, Jónasson & Rist,
1992). Frostastaðavatn is a smaller (2.55 km2), shallower (mean depth = 6.5 m) lake located in the highlands of southern Iceland (64∞03¢N, 19∞00¢W).
G. aculeatus were collected from both the littoral lava
habitat and the soft-bottom habitat of each of these
lakes. Specimens were also collected from a third lake,
Sauravatn, a small (0.84 km2), very shallow (mean
depth = 0.2 m) wetland lake located in the M y¢ rar district in south-west Iceland (64∞40¢N, 22∞10¢W). Sauravatn has a uniform, vegetated benthic habitat, and
contains only a monomorphic G. aculeatus population
with no piscine predators. Two marine sites were
included in the experimental design, and it was
assumed that they resembled the original ancestral
population of G. aculeatus prior to their immigration
to freshwater systems. Stokkseyri is on the south
shore of Iceland near the estuary of the River Ölfusá
(63∞50¢N, 21∞04¢W), which drains Thingvallavatn via
the River Sog. Hvassahraun is on the south-western
coast of Iceland (64∞01¢N, 22∞08¢W).
G. aculeatus were usually collected using wire minnow traps (No. 12562, Canada Fishing and Tackle
Sports Ltd; mesh size = 3.2 mm). The traps were
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
DIVERGENCE OF STICKLEBACKS
191
Table 1. Description of the seven sites where Gasterosteus aculeatus were collected
Site
Predation risk
Depth (m) Vegetation
Special Features
Thingvallavatn lava
(TLAVA)
High – piscivorous
Salv. alpinus & Salmo
trutta
1
Thingvallavatn
soft-bottom
(TSOFT)
Very high – piscivorous
Salv. alpinus & Salmo
trutta
17
Frostastaðavatn
lava (FLAVA)
Moderate – recently
introduced Salv. alpinus
& Salmo trutta
Frostastaðavatn
soft-bottom
(FSOFT)
Moderate – recently
introduced Salv. alpinus
& Salmo trutta
Sauravatn (SAUR)
(Control)
Minimal – no piscine
predators
<0.5
Dense; multiple
species
Very shallow, marshy lake;
G. aculeatus collected in vegetation at shoreline
Stokkseyri (STOKK)
(Marine)
Unknown – presumed
minimal
<1
Fucus sp. & dense
blue-green algae
Isolated marine tide pools on
lava shoreline
Hvassahraun (HVASS)
(Marine)
Unknown – presumed
moderate
2
Sparse Fucus sp. &
other marine
seaweeds
Marine inlet on lava shoreline
Patchy green algae
Lava rubble with subbenthic
(e.g. Tetraspora sp.)
spaces; cold, stable spring
inflow
Nitella opaca (dense
stands to height of
1 m)
No lava; N. opaca stands contain
dense G. aculeatus populations
surrounded by open water
5
Minimal
Steep lava shoreline; very dense
G. aculeatus population
1–1.5
No lava; shallow, sandy shoreNitella sp. common,
line; very dense G. aculeatus
Chara sp. & Myriopopulation
phyllum sp. patchy
weighted, attached by line to surface buoys, and left in
the water for approximately 24 h. Attempts to catch
G. aculeatus with traps at the Thingvallavatn lava site
were unsuccessful and they were collected at this site
with the use of electro-fishing gear.
G. aculeatus were maintained in the field in 100-L
containers aerated with battery-operated air pumps
for up to 5 days and they were fed dried Tubifex worms
daily. The containers were kept outdoors in a shaded
area. Outdoor temperatures ranged from 10 to 15∞C.
The fish were taken to the freshwater aquarium of
Hólar College, Iceland and immediately transferred to
80-L glass aquaria (L ¥ W ¥ H = 50 ¥ 40 ¥ 40 cm; flowthrough water 2 L/min; temperature 10∞C ± 1∞C; photoperiod 16 : 8 h light : dark). G. aculeatus from each
site were maintained separately and 45 adults were
placed in each tank for approximately 4 weeks prior to
the start of the experiments. They were fed daily dried
Tubifex, frozen or dried chironomid larvae, or flake
food. All G. aculeatus were kept under exactly the
same conditions; any effects of tank enclosure would
be the same for all populations.
In the predation experiments, two brown trout
Salmo trutta (approximately 25 cm standard length)
from a small local stream (65∞44¢N, 19∞05¢W) and one
Salv. alpinus (approximately 25 cm standard length)
from the local lake Vatnshlídarvatn (65∞31¢N,
19∞37¢W) were used as piscine predators. These three
fish had been maintained for approximately 1 year in
large 800-L tanks with other members of their respective populations in the freshwater aquarium at Hólar.
The predators were fed aquaculture feed daily and
had no experience of G. aculeatus while in captivity.
SHOALING
DENSITY METHODOLOGY
Experimental trials were conducted in a large flowthrough aquarium (L ¥ W ¥ H = 160 ¥ 70 ¥ 80 cm;
water temperature 10∞C ± 1∞C; Fig. 1). The back and
sides of the tank were covered with white plastic to
minimize external disturbance and to make
G. aculeatus more visible. A Plexiglas divider was
placed 40 cm from one end of the aquarium (Fig. 1).
Three rows of holes (3 mm in diameter) along the
edges of the Plexiglas divider allowed the free flow of
water through the divider and the passage of odours
from the predators and G. aculeatus. A large mirror
(L ¥ W = 160 ¥ 90 cm) was angled above the tank to
permit three-dimensional observations. Markings on
the front of the glass aquarium and strings across its
top divided the aquarium into 96 cubes
(L ¥ W ¥ H = 20 ¥ 20 ¥ 17.5 cm or 7000 cm3). The size
of the grid was selected so as to represent approximately four G. aculeatus body lengths (4 ¥ 5 cm). It is
within this proximity of conspecifics that fish can be
considered to be in a shoal (Pitcher & Parrish, 1993).
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
192
L. I. DOUCETTE ET AL.
Mirror
Divider
80 cm
70 cm
40 cm
120 cm
Figure 1. Experimental set-up for shoaling experiments. In the first set of trials, no predatory fish were present to the
left of the transparent divider. In the second set of trials, two brown trout (Salmo trutta) and one arctic charr (Salvelinus
alpinus) were present in the compartment to the left of the divider.
Table 2. Mean standard length* of Gasterosteus aculeatus for each shoaling density replicate (N = 22 for all populations)
TLAVA
TSOFT
FLAVA
FSOFT
SAUR
STOKK
HVASS
Replicate 1
Length (mm)
Standard error
48.2
1.44
51.5
0.62
69.1
1.19
61.3
1.20
47.0
0.87
52.5
0.49
60.1
1.24
Replicate 2
Length (mm)
Standard error
40.5
1.50
47.4
0.75
57.7
0.82
56.6
1.25
41.4
0.83
51.4
0.62
54.5
0.87
*Mean standard length measured to nearest 0.5 mm. Site abbreviations are defined in Table 1.
Lighting consisted of three 60-W incandescent lamps
illuminating the tank from above and a 125-W halogen spotlight illuminating the tank from the front. No
cover was provided in the tank for G. aculeatus
because we wanted to determine how they would react
to predators in an open-water situation, where they
would be unable to seek refuge.
In the wild, fish discriminate between potential
shoal mates on the basis of features such as body size
(Ranta, Juvonen & Peuhkuri, 1992a; Ranta, Lindstrom & Peuhkuri, 1992b; Ward & Krause, 2001) and
parasite load (Barber & Huntingford, 1995; Barber,
Huntingford & Crompton, 1995). Only healthy, non-
parasitized G. aculeatus were selected for these experiments. Two replicate groups of 22 G. aculeatus from
each site were tested. Given that fish typically form
shoals based on body size, the first replicate group consisted of the largest G. aculeatus from the population
sample and the second group consisted of G. aculeatus
of a slightly smaller size (Table 2). They were tested in
groups by site. Gender is not typically considered in
studies of shoaling (e.g. Pitcher & Parrish, 1993;
McRobert & Bradner, 1998; Ward & Krause, 2001),
and gender differences were not considered in this
study. Our experiments were conducted outside of the
breeding season for G. aculeatus in Iceland, so any
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
DIVERGENCE OF STICKLEBACKS
gender effects would be minimal. Furthermore,
because of the monomorphic external appearance of
the two genders when not in breeding colours, a fish’s
gender could not be determined without dissection.
In initial trials predators were absent, and
G. aculeatus were transferred directly from the holding tanks to the test aquarium. They were allowed
15 min to acclimatize to the tank before recording
began. Two video cameras, one directed at the front of
the tank and the other at the mirror angled above the
tank, were used to record the behaviour and position
of the group for 20 min. The fish were returned to their
holding tanks at the end of the 20-min trial.
Twenty-four hours after each initial shoaling trial
(with predators absent), the same groups of
G. aculeatus were retested in a second trial, in which
the three predators were positioned in the smaller section (L ¥ W ¥ H = 40 ¥ 70 ¥ 80 cm) of the aquarium
prior to the release of G. aculeatus into the larger section. We used live predators because previous studies
have shown that fish habituate more slowly to real
predators than they do to models (Pitcher, 1992).
Because the predators were already present and
G. aculeatus were familiar with the tank, the procedure for introducing G. aculeatus to the test tank was
altered in this second trial. They were placed in a
Plexiglas box, with a bottom of mesh screening, and
held in the water at the far side of the tank for 1 min
prior to release in the centre of the tank. This allowed
them to recover from the transfer and adjust to the
lighting without direct observation of the predators.
They showed no response to the presence of the predators prior to release from the box.
The videotapes were later analysed for shoaling
density by pausing the tape every 30 s throughout the
first 15 min of each predator-absent and predatorpresent trial and recording the three-dimensional
position of each G. aculeatus in the grid pattern (i.e.
instantaneous scan sampling). The video recordings
had to be forwarded second-by-second and reversed
regularly to ensure the accurate position of all 22 fish
in three-dimensional space. Only the first 15 min of
the trials were used due to the time required to complete this task, yielding 30 instantaneous measures of
shoaling density for each of the two replicates.
A ‘shoal’ is a group of fish that remain together due
to social attraction (Pitcher & Parrish, 1993). A ‘school’
refers to a synchronized and cohesive group that is orientated and moving together in the same direction
(polarized), making it a subcategory in the broader
term of shoaling (Pitcher & Parrish, 1993). In this
paper, shoaling density/cohesion was examined using
a grid to determine the spatial organization of a group
of G. aculeatus; schooling was examined separately as
antipredator behaviour that occurs in the presence of
a predation threat.
OBSERVATIONS
193
OF PREDATOR INSPECTION AND
AVOIDANCE BEHAVIOUR
The video recordings from the predator-present trials were analysed further for inspection and avoidance behaviour related to the presence of predators.
Inspection, schooling, fleeing and startle responses
were recorded from the front view (behaviour categories are defined in Table 3), using The Observer
(Noldus, Netherlands, Version 3) behaviour recording software. States and events were recorded for
the entire duration of the 20-min trial. Each trial
analysis required that the recording be viewed several times for each event or state to be recorded
accurately. Because of the extensive time required,
and the observation that both replicates from each
of the populations showed very similar behaviour,
only one trial was analysed for each G. aculeatus
population.
Movements of the predatory fish were recorded
throughout the trials. Reactions to predator movements differed between groups and no direct correlation was found. For instance, a charge at the
divider by a predator would elicit startle and flee
responses in one group of G. aculeatus but no
reaction in another. Furthermore, some groups
showed frequent startle and flee responses when
predators were stationary or facing away from
them.
Table 3. Ethogram for antipredator behaviour of Gasterosteus aculeatus
States
Inspection – one or more G. aculeatus leave the main
shoal and slowly approach the predator, remain still in the
water column facing the predator, and later return to the
shoal; recorded from time inspecting individuals approach
within 50 cm of the predator until they turn to return to
the main shoal; recorded for groups of 1, 2, 3, 4, 5, 6, 7–9,
and ≥10 G. aculeatus
Schooling – groups of three or more G. aculeatus that are
swimming in a synchronized and polarized group; recorded
for groups of 3, 4, 5, 6–10, 11–15, and ≥16 G. aculeatus
Events
Startle response – flexing of the body laterally (C-shape)
combined with a quick release which results in a fast movement of the G. aculeatus away from a perceived threat;
sometimes followed by fleeing; recorded for groups of 1, 2,
3, 4, 5 and >5 G. aculeatus
Flee response – a sudden swimming dash in a direction
away from a predator or a perceived threat; recorded for
groups of 1, 2, 3, 4, 5, 6, and >6 G. aculeatus
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
194
L. I. DOUCETTE ET AL.
STATISTICAL
ANALYSIS OF SHOALING DENSITY
To determine shoaling density, frequency histograms
for both predator-absent and predator-present trials
were created to show the frequency in which groups of
1–22 G. aculeatus were recorded in each three-dimensional cube (see Appendix 1, Doucette, 2001). No differences in shoaling density were found between the
two replicates for each of the seven populations
(Mann–Whitney (M–W) U-tests: N = 22, P > 0.05).
Therefore, the data for the two replicates were pooled
to create a two-way table of responses for the frequency of groups of £5 and >5 G. aculeatus occurring
within a single grid cube, and the results were plotted
as bar graphs (Fig. 2; Koopmans, 1987). The presence
of >5 G. aculeatus within a grid cube was indicative of
a tightly formed shoal.
To further examine the shoaling density, an ‘index of
cohesion’ was calculated (Seghers, 1974). At each 30-s
A
Stickleback group
HVASS
STOKK
SAUR
FSOFT
FLAVA
TSOFT
TLAVA
0
10
20
30
40
50
60
70
80
90
100
Relative frequency (%)
B
Stickleback group
HVASS
STOKK
SAUR
FSOFT
observation, the maximum density for any of the 96
grids was recorded. The index is the mean maximum
density for 30 consecutive observations in each of two
separate replicates. It has a theoretical minimum of
one (all G. aculeatus found singly) and a maximum of
22 (entire group in the area of one cube). Kruskal–
Wallis (K–W) and M–W U-tests were used to compare
the differences among populations for mean maximum
shoal density (Sokal & Rohlf, 1994).
STATISTICAL ANALYSIS OF PREDATOR INSPECTION AND
AVOIDANCE BEHAVIOUR
Inspection and schooling data were examined to determine: (1) the intensity of the respective behaviour by
calculating the total duration for which an inspection
unit or school existed during the trials, and (2) the cost
to the individual G. aculeatus by calculating the duration of time that an average individual spent in an
inspection unit or school. G. aculeatus were classified
as an inspecting unit as long as they were all inspecting and responding to the predators synchronously.
They were classified as schooling when they were moving as in synchronized and polarized formation. The
total duration of inspection and schooling behaviour
was calculated and compared for the seven
G. aculeatus groups (K–W test) and specifically for the
sympatric groups within lakes (M–W U-test). To determine the cost of inspection and schooling for an average individual G. aculeatus, the number of fish
demonstrating that behaviour was multiplied by the
duration of the behaviour. For the purposes of statistical analysis, the duration of a behaviour occurring
within each minute of the 20-min trial was treated as
a subsequent observation or bout (Lehner, 1996). The
total number of startle and flee responses was
calculated for each group and compared using K–W
and M–W U-tests.
FLAVA
RESULTS
TSOFT
SHOALING
TLAVA
0
10
20
30
40
50
60
70
80
90
100
Relative frequency (%)
Figure 2. Results of the two-way table of responses showing the relative frequency distribution (%) of groups of £5
Gasterosteus aculeatus (black bars) or groups with >5
G. aculeatus (white bars) being present in a single grid
cube (20 ¥ 20 ¥ 17.5 cm) during the duration of the trial for
(A) predators absent and (B) predators present. Perpendicular bars represent the mean relative frequency (%) of
shoals of £5 Gasterosteus aculeatus (black bar) or >5
G. aculeatus (white bar) for all seven G. aculeatus groups
combined. Site abbreviations are defined in Table 1.
DENSITY AND COHESION
Gasterosteus aculeatus from both TSOFT and HVASS
sites shoaled in groups >5 more in comparison with
the other sites when predators were absent (Fig. 2A).
The combined mean of all groups revealed that shoals
containing >5 G. aculeatus were formed only 23% of
the time. However, the TSOFT group formed shoals of
>5 G. aculeatus 49% of the time and the HVASS group
formed these larger shoals 71% of the time.
G. aculeatus from HVASS showed the highest shoaling
affinity, with 53% of HVASS G. aculeatus recorded in
shoals of ≥10 and only 8% of the fish recorded singly.
The TSOFT group formed shoals of ≥10 G. aculeatus
34% of the time. In contrast, FSOFT and SAUR groups
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
DIVERGENCE OF STICKLEBACKS
rarely formed shoals of >5 G. aculeatus, and the
FLAVA group never formed such shoals (Fig. 2A).
When predators were present there were marked
changes in shoaling density. G. aculeatus from TSOFT
increased their shoaling density, forming shoals >5
64% of the time (Fig. 2B) and, forming shoals >10 49%
of the time. TSOFT G. aculeatus were found singly for
only 10% of the predation trial. Similarly, FLAVA,
FSOFT and SAUR groups all increased the number of
large shoals formed in the presence of predators
(Fig. 2B). Contrary to what one might predict, the
HVASS group decreased their shoaling density and
tended to form shoals of £5 G. aculeatus 81% of the
time vs. 29% when predators were absent. TLAVA and
STOKK groups also decreased the number of large
shoals formed (13% vs. 19% and 3% vs. 20%, respectively) when predators were present.
Comparisons between the two replicate trials for
both predators absent and predators present revealed
that in some cases the second group tested, consisting
of slightly smaller fish, showed higher shoaling cohesion (Table 4). Juveniles of most fish species are
known to form cohesive shoals (Pitcher & Parrish,
1993). While all G. aculeatus tested in these experiments were adults, it is possible that the smaller fish
(second trials) were slightly younger and less experienced. These fish would achieve greater security in
larger and more compact shoals (Seghers, 1974;
Magurran, 1990). The only groups to show significantly greater shoal cohesion for the second trial with
and without predators present were FSOFT and
SAUR. Generally, shoaling cohesion for both of these
groups was very low, thus this difference between trials had little impact on the overall results.
There was a highly significant difference in the
mean maximum shoal density among all groups both
when predators were absent (K–W test: H6 = 275.8,
P < 0.00001) and when predators were present (K–W
195
test: H6 = 191.1, P < 0.00001). As expected, the index
of cohesion revealed similar trends as in the analysis
of shoals (Fig. 3). When predators were present a
significant increase in mean maximum density
was seen in three of the groups: TSOFT (M–W Utest: U = 2359.5, N1 = N2 = 60, P < 0.005), FLAVA
(U = 2838.5, N1 = N2 = 60, P < 0.00001), and FSOFT
(U = 2975, N1 = N2 = 60, P < 0.00001). However, the
mean maximum density of HVASS shoals decreased
from 13.3 to 6.1 G. aculeatus when predators
were present (M–W U-test: U = 377, N1 = N2 = 60,
P < 0.00001).
PREDATOR
INSPECTION AND AVOIDANCE BEHAVIOUR
We concluded that in the HVASS and TLAVA groups it
was the high incidence of inspection behaviour that
led to decreased shoal size when predators were
present. The continuous movement of one or more
G. aculeatus away from the main shoal to inspect
caused a decrease in shoal density. All groups had one
or more G. aculeatus inspecting during at least 76% of
the total trial duration (Table 5). There was a significant difference between groups for the total duration
of time that an inspection unit was present (K–W:
H6 = 29.99, P < 0.00005; Table 5). The TLAVA and
SAUR groups had longer inspection durations than
did the other groups (Dunn’s multiple comparison,
P < 0.05). When the duration of units of £5
G. aculeatus and >5 G. aculeatus inspecting were
examined separately, the differences between groups
was even greater (K–W: H6 = 41.04, P < 0.00001 and
H6 = 41.88, P < 0.00001, respectively). These differences were caused by the tendency for TLAVA
G. aculeatus to inspect in much larger units than did
all other populations (Table 5). The number of inspection events per trial for TLAVA G. aculeatus was less
than it was in the other groups, indicating that these
Table 4. Mean maximum shoal density (cohesion) and results of Mann–Whitney U-tests for seven G. aculeatus populations comparing the two replicate trials (N1 = N2 = 30 for all populations)
TLAVA
TSOFT
FLAVA
FSOFT
SAUR
STOKK
HVASS
Predators absent
Mean Replicate 1
Mean Replicate 2
U-value
P-value <0.05
3.10
7.87
879.5
Yes
10.13
10.93
500
No
2.13
2.03
420
No
2.30
3.30
661
Yes
2.50
3.70
671
Yes
3.03
3.13
492
No
13.73
12.87
399.5
No
Predators present
Mean Replicate 1
Mean Replicate 2
U-value
P-value <0.05
4.93
5.50
520
No
14.60
10.90
178.5
Yes
3.20
4.23
518
No
3.47
7.40
862
Yes
3.20
3.97
594.5
Yes
3.83
3.57
419
No
5.13
7.10
603.5
Yes
Site abbreviations are defined in Table 1.
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
196
L. I. DOUCETTE ET AL.
A
Index of cohesion
20
15
10
5
H
H
VA
SS
VA
SS
KK
ST
ST
O
KK
O
R
U
SA
FS
O
FT
A
AV
FL
TL
TS
O
AV
A
FT
0
Stickleback group
Index of cohesion
20
B
15
10
5
R
U
SA
FT
FS
O
A
AV
FL
FT
O
TS
TL
AV
A
0
Stickleback group
Figure 3. Index of cohesion showing the mean maximum
shoal density for seven groups of Gasterosteus aculeatus.
Boxes represent the means of two replicates and 60 instantaneous samples per group. Error bars represent ±95%
confidence intervals and dashes represent the maximum
and minimum values of maximum shoal density for each
group. (A) Predators absent. (B) Predators present. Site
abbreviations are defined in Table 1.
large units continued inspecting for long periods of
time before returning to the main shoal. There were
clear differences in inspection duration between the
two groups from Thingvallavatn. The total time that
TLAVA G. aculeatus inspected was significantly
greater (M–W U-test: U = 203, N1 = N2 = 20, P < 0.05)
and the time that TLAVA G. aculeatus inspected in
units >5 was longer than it was for the TSOFT
G. aculeatus (M–W U-test: U = 364, N1 = N2 = 20,
P < 0.00001). Total inspection duration did not differ
between the two groups in Frostastaðavatn (M–W Utest: U = 145, N1 = N2 = 20, P = 0.14).
Based on the duration of inspection and the number
of G. aculeatus in each inspection unit, the duration
that an individual inspected was calculated (Pitcher,
1992). There were significant differences among
groups in the total duration that each G. aculeatus
performed inspection behaviour (K–W: H6 = 50.19,
P < 0.00001; Fig. 4). These differences were due
mainly to the much longer duration of individual
inspection by the TLAVA G. aculeatus (Dunn’s multiple comparison test: P < 0.05). An individual in the
TLAVA population spent an average of 64% of its time
inspecting compared with a maximum of 25% (SAUR)
in other groups, and it was almost always in a group of
>5 G. aculeatus. There was a large difference in total
individual inspection duration between the two
groups in Thingvallavatn (M–W U-test: U = 400.0,
N1 = N2 = 20, P < 0.00001), but no significant differences between the two groups in Frostastaðavatn
(M–W U-test: U = 177.0, N1 = N2 = 20, P = 0.54).
The duration for which a school was present during
the predation trials varied greatly among the groups
(K–W: H6 = 32.79, P < 0.00001; Table 5). As with
shoaling behaviour, HVASS G. aculeatus showed the
greatest duration of schooling (41%) followed by
TSOFT G. aculeatus (24%). The two marine groups
(STOKK and HVASS) and TSOFT G. aculeatus spent
significantly more time schooling than did the other
groups (Dunn’s multiple comparison test: P < 0.05).
The marine populations also tended to form larger
schools both in terms of size at first school and most
frequent school size (Table 5). SAUR G. aculeatus only
schooled for 1% of the total trial time. There were no
significant differences in total school duration
between the two groups in Thingvallavatn (M–W Utest: U = 136.5, N1 = N2 = 20, P = 0.09) or in Frostastaðavatn (M–W U-test: U = 210, N1 = N2 = 20,
P = 0.80). However, the TSOFT group formed more
small schools of £5 G. aculeatus than did the TLAVA
group (M–W U-test: U = 90.5, N1 = N2 = 20, P < 0.005).
Individuals from the two marine populations spent
more time schooling than did individuals from other
groups (K–W: H6 = 31.66, P < 0.00005; Fig. 5). In most
groups, individuals tended to be in schools of >5
G. aculeatus. However, TSOFT G. aculeatus formed
small schools of £5 more often than did the other
groups (K–W: H6 = 43.67, P < 0.00001; Dunn’s multiple comparison test: P < 0.05). There were no significant differences in the total duration for which an
individual schooled between the sympatric groups in
Thingvallavatn (M–W U-test: U = 151.5, N1 = N2 = 20,
P = 0.19) nor between the groups in Frostastaðavatn
(M–W U-test: U = 216, N1 = N2 = 20, P = 0.67).
Differences among groups in numbers of startle and
flee responses was highly significant (K–W:
H6 = 43.28, P < 0.00001 and H6 = 60.70, P < 0.00001,
respectively). The TLAVA G. aculeatus executed a
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
DIVERGENCE OF STICKLEBACKS
197
Table 5. Summary of inspection and schooling data from shoaling trials with predators present
TLAVA
TSOFT
FLAVA
FSOFT
SAUR
STOKK
HVASS
Inspection
Total duration inspection occurs (%)
Inspection unit £5 (%)
Inspection unit >5 (%)
Latency to first inspection (≥1) (s)
Size of first inspection unit (N)
Number of inspection events
Most frequent inspection unit size
Inspection unit size of longest duration
98.10
7.11
90.99
0.0
2
51
≥10
≥10
82.76
44.50
38.26
0.0
3
113
1&2
7–9
88.63
55.19
33.43
3.0
1
139
1&2
≥10
81.10
48.90
32.20
0.0
≥10
152
1
7–9
97.96
58.43
39.53
0.0
3
105
4
7–9
76.06
44.16
31.90
0.0
1
112
1
1
88.10
50.28
37.28
0.0
≥10
159
2
2
Schooling
Total duration a school is present (%)
School £5 (%)
School >5 (%)
Latency to first school (≥3) (s)
Size of first school (N)
Number of schooling events
Most frequent school size
School size of longest duration
12.38
1.45
10.93
0.0
6–10
22
6–10
≥16
24.30
14.30
10.00
0.0
4
38
6–10
6–10
7.64
4.15
3.49
33.5
≥16
13
4
4
9.84
1.03
8.81
26.0
5
15
6–10
11–15
1.04
0.00
1.04
4.0
6–10
2
6–10
6–10
23.84
0.63
23.21
4.0
≥16
30
11–15
≥16
40.46
10.17
30.29
0.0
≥16
44
11–15
≥16
Data excluding ‘0’, no school or inspection unit present. Site abbreviations are defined in Table 1.
25
70
60
20
Total duration (%)
Total duration (%)
50
40
30
15
10
20
5
10
0
TLAVA TSOFT FLAVA FSOFT SAUR
Stickleback group
STOKK HVASS
0
TLAVA TSOFT
FLAVA FSOFT SAUR
Stickleback group
STOKK HVASS
Figure 4. Total duration (%) that an average individual
Gasterosteus aculeatus spent inspecting during a 20-min
trial in the presence of predators. Calculations were based
on the total inspection duration ¥ number of G. aculeatus
present in each inspection unit. Site abbreviations are
defined in Table 1. = inspection unit £5; = inspection
unit >5; = total inspection behaviour.
Figure 5. Total duration (%) that an average individual
Gasterosteus aculeatus spent schooling during a 20-min
trial in the presence of predators. Calculations were based
on the total duration a school was present ¥ number of
G. aculeatus present in each school. Site abbreviations
are defined in Table 1. = school £5 G. aculeatus;
=
school >5 G. aculeatus; = total schooling.
larger number of startle and flee responses than did
all other groups (Dunn’s multiple comparison test:
P < 0.05; Fig. 6). There was no significant difference in
the number of startle responses between the two
groups in Thingvallavatn (M–W U-test: U = 146,
N1 = N2 = 20, P = 0.15) but there was a clear difference
in the number of fleeing manoeuvres (M–W U-test:
U = 301, N1 = N2 = 20, P < 0.05). There was a difference in both the number of startle responses (M–W Utest: U = 300, N1 = N2 = 20, P < 0.05) and the number
of fleeing manoeuvres (M–W U-test: U = 290,
N1 = N2 = 20, P < 0.05) in Frostastaðavatn.
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
198
L. I. DOUCETTE ET AL.
Total no. startle + flee responses
160
140
120
100
80
60
40
20
0
TLAVA
TSOFT
FLAVA FSOFT SAUR
Stickleback group
STOKK HVASS
Figure 6. Total numbers of startle () and flee ()
responses during a 20-min trial in the presence of predators for seven populations of Gasterosteus aculeatus. Site
abbreviations are defined in Table 1.
DISCUSSION
POPULATION
DIFFERENCES
The results of this study show clear differences in
antipredator behaviour, including shoaling, schooling,
and startle and flee responses, between G. aculeatus
populations residing in different habitats. These differences appear to relate to the apparent risk of predation. G. aculeatus from Thingvallavatn, where the
risk of predation is very high (Malmquist et al., 1992),
showed the most vigorous antipredator responses. The
groups from Thingvallavatn also exhibit the greatest
morphological differences compared with G. aculeatus
pairs in other volcanic Icelandic lakes (Kristjánsson,
2001; Kristjánsson et al., 2002a) and show clear differences in other behaviour (Doucette, 2001). In contrast, in Frostastaðavatn, where the risk of predation
may be lower, morphological (Kristjánsson et al.,
2002a) and behavioural divergence (Doucette, 2001) of
G. aculeatus is less. Despite the presence of piscine
predators, neither Frostastaðavatn group showed
well-developed antipredator responses, and the shoaling cohesion of both groups was very low. G. aculeatus
from Sauravatn, where no piscine predators exist,
showed almost no antipredator behaviour. Marine
populations relied almost entirely on shoaling and
schooling to deter predators. Freshwater populations
appear to have lost some affinity for forming the dense
shoals or schools seen in marine populations, perhaps
in favour of seeking refuge when confronted by a
potential predator. Results suggest that the risk of
predation may be an important contributing factor in
promoting ecological divergence and maintaining separate forms. Under the right circumstances, predation
pressure could force prey to compete for predator-free
resource habitats and lead to sympatric speciation.
One morph of Salv. alpinus in Thingvallavatn is
specialized to feed on G. aculeatus (Sandlund et al.,
1987; Malmquist et al., 1992). Based on diet analysis
of Salv. alpinus in the littoral lava habitat, between
35% and 100% of the stomachs of the large limnetic
piscivore morph contained G. aculeatus (diet varied by
season). Between 65% and 100% of the ash-free dry
weight (AFDW) of the large limnetic morph’s diet consisted of G. aculeatus. In the profundal and littoral
Nitella (soft-bottom) habitats 44–100% of the large
limnetic morph stomachs contained G. aculeatus
(AFDW = 36–88%). The present habitat structure of
the lake was already in place shortly after the ice
retreated from the area around 10 000 years ago
(Sæmundsson, 1992). Salv. alpinus, Salmo trutta and
G. aculeatus most probably colonized the lake quickly
after this. The presence of a specialist piscivorous
Salv. alpinus in Thingvallavatn (Sandlund et al.,
1987; Malmquist et al., 1992) suggests that predation
pressure on G. aculeatus has been intense for thousands of years.
While both Thingvallavatn groups showed a clear
reaction to the predators, their responses were vastly
different. The TSOFT population in Thingvallavatn
shoaled cohesively, a suitable response in a comparatively open, vegetated environment. It is well known
that small prey fish reduce individual risk of predation while shoaling and will form larger shoals in
high-risk localities, even in the absence of direct
threat, as vigilance increases with shoal size (Seghers,
1974; Magurran & Seghers, 1994a; Magurran, 1999).
Typically, when a potential underwater predator is
detected, shoals become more compact and cohesive
(Pitcher & Parrish, 1993), as was the case for the
TSOFT population, which formed polarized schools for
24% of the duration of the trial. In contrast, the
TLAVA G. aculeatus from Thingvallavatn showed only
moderate shoaling and schooling behaviour, likely due
to the availability of sub-benthic spaces in their habitat that provide safe cover from large predatory fish.
The bottom of the Vatnsvik inlet, where the TLAVA
G. aculeatus were collected, is composed of broken
lava rock with sub-benthic spaces and rifts, which may
reach depths of more than 10 m (Sæmundsson, 1992).
The spaces between and under rocks provide small
fish with a protected environment (Snorrason et al.,
1994). G. aculeatus are rarely observed in the open
water above the lava rubble. The intense monitoring
and inspection behaviour may enable the TLAVA
G. aculeatus to evade predation by early detection of
motivated predators. The benefit of inspection is that
it provides continually updated information about
predator identity and likelihood of attack (Pitcher,
1992). This reduces the risk to all individuals by allowing attack anticipation. When the perceived risk is
high, risk dilution or ‘safety in numbers’ is achieved by
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
DIVERGENCE OF STICKLEBACKS
forming larger inspection units (Magurran, 1990;
Pitcher, 1992; Magurran & Seghers, 1994a). In this
study, inspection units of TLAVA G. aculeatus often
approached the maximum shoal size of 22 (see Magurran & Pitcher, 1987) whereas the predominant inspection unit size for G. aculeatus from all other sites was
one or two. The TLAVA G. aculeatus coordinated their
lengthy inspection bouts with the frequent use of startle and flee responses. Despite the variability in antipredator behaviour in the two groups from
Thingvallavatn, both showed appropriate responses to
a high-risk predatory habitat.
The antipredator responses observed for both
groups of G. aculeatus in Frostastaðavatn were minimal. Both groups showed low levels of shoaling and
schooling behaviour. Based on lake structure, FSOFT
G. aculeatus would be expected to show a high degree
of schooling and shoaling behaviour, similar to the
intense schooling behaviour of TSOFT G. aculeatus.
Only the relatively high number of startle and flee
responses of the FSOFT G. aculeatus suggests that
these fish are from a lake with a dense population of
potential piscine predators. In Frostastaðavatn,
G. aculeatus have most likely evolved without threat
from piscine predators until Salv. alpinus and Salmo
trutta were introduced about 30 years ago (B. Jónsson,
pers. comm.). The time passed since the introduction
of Salv. alpinus equals approximately 15 G. aculeatus
generations, which is probably too short to cause a
significant evolutionary response. Furthermore,
Salv. alpinus in Frostastaðavatn may not be predisposed to feed on G. aculeatus. Stomach content analysis indicated that only 4% of their diet was composed
of G. aculeatus (B. Jónsson, unpubl. data).
The lack of predation threat in Sauravatn has led to
a paucity of antipredator behaviour in SAUR
G. aculeatus. Shoaling cohesion was extremely low
and increased only minimally when predators were
present. These fish showed almost no schooling behaviour, and inspection was equal to or less than that in
all other groups. Startle and flee responses were also
rare. The presence of piscine predators during the
experiments did not elicit behavioural responses considered to reduce predation risk. The inspection
behaviour that was observed was perhaps due more to
curiosity than fear. This population of G. aculeatus
represents an excellent control group for antipredator
behaviour, enforcing the prediction that low natural
antipredator behaviour would exist in G. aculeatus
from a lake with no risk of piscine predation.
The marine G. aculeatus populations were selected
as examples of the ancestral form of G. aculeatus in
Iceland for comparison with the evolved freshwater
populations (Bell & Foster, 1994a). The marine populations showed dense shoaling and schooling as their
main means of antipredator defence. Freshwater pop-
199
ulations generally appear to have lost some affinity for
forming these dense schools and preferred to utilize
escape manoeuvres. HVASS G. aculeatus showed the
greatest frequency of shoaling in large groups and the
greatest cohesion. In some instances the shoaling density of HVASS G. aculeatus was so great that >19
G. aculeatus occupied one small cube in the tank
(7000 cm3). The degree of shoaling was so intense in
the absence of predators that shoaling decreased significantly when predators were present because small
groups left the shoal to inspect the predators.
Surprisingly, the STOKK population showed relatively low levels of shoaling, but like the HVASS
G. aculeatus, they formed schools for long durations.
Schooling can be seen as a measure of the impact of
predation risk on fish travelling through the water column (Pitcher & Parrish, 1993). The fact that STOKK
G. aculeatus school in the presence of predators, but
do not shoal regularly, indicates an awareness of predation risk when in open water. Compared with
HVASS G. aculeatus, STOKK G. aculeatus showed a
relatively large number of startle and flee responses.
Perhaps this is a response to spending the breeding
season (late May to late August) in a tide-pool habitat
where shelter is readily available but the formation of
large shoals is relatively difficult.
Findings by Taylor & McPhail (1986) suggest that
evasive manoeuvres, including startle and flee
responses, may not be the most efficient means of
defence for marine G. aculeatus populations. Marine
G. aculeatus have been found to be better at prolonged swimming and poorer at burst swimming
than are freshwater G. aculeatus (Taylor & McPhail,
1986). Body flexion and maximum velocity attained
during startle responses was greater in resident
freshwater populations than it was in marine
G. aculeatus. This is thought to reflect a trade-off
between swimming abilities and defensive armour
(Andraso, 1997). It is known that marine
G. aculeatus generally have a greater number of
armour plates than do freshwater G. aculeatus
(Reimchen, 1983; Bell & Foster, 1994a). Morphological studies of Icelandic G. aculeatus are in accordance with this (G. Ólafsdóttir, unpubl. data). Both
HVASS and STOKK G. aculeatus were observed to
be full-plated morphs (Bell & Foster, 1994b) with a
greater number of armour plates than the freshwater populations (L. I. Doucette, unpubl. data). The
change from schooling and heavy body armour to
increasing flexibility and evasive manoeuvres following the settlement in freshwater systems is most
probably caused by the greatly increased possibilities to seek effective shelter in freshwater systems
(see Introduction; Taylor & McPhail, 1986;
Klepacker, 1993; Andraso & Barron, 1995; Andraso,
1997).
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
200
RISK
L. I. DOUCETTE ET AL.
OF PREDATION PROMOTES SPECIES DIVERGENCE
Ecological stability, discreteness and heterogeneity in
niches, and low fish-species diversity are seen as relevant features promoting phenotypic divergence
(Maynard Smith, 1966; Malmquist et al., 1992; Skúlason et al., 1999; Snorrason & Skúlason, 2004). Both
Thingvallavatn and Frostastaðavatn are in areas of
permeable bedrock and each lake is fed primarily
through underground springs (Adalsteinsson et al.,
1992; B. Jónsson, pers. comm.). Thus, the waters in
both lakes are rich in minerals, and are generally stable and productive (Jónasson, Adalsteinsson & Jónsson, 1992.). Both lakes contain only three species of
fish, hence there is low interspecific competition, and
have relatively similar lava and soft-bottom habitats.
So why is there not the same level of morphological
(Kristjánsson et al., 2002a) and behavioural (Doucette, 2001) divergence between G. aculeatus occupying
the two diverse habitats in each lake? When considering the variable degree of divergence of sympatric
groups of G. aculeatus in Thingvallavatn and Frostastaðavatn we must examine several ecological factors
that differ between the lakes and which may influence
the probability of morph divergence: (i) the lake size,
(ii) the evolutionary time available for habitat-specific
divergence, and (iii) the level of predation and the
proximity of alternative G. aculeatus habitats.
Thingvallavatn has an area of 83 km2 and a mean
depth of 34 m (Adalsteinsson et al., 1992) whereas
Frostastaðavatn is only 2.55 km2 and has a mean
depth of 6.5 m (H. Malmquist & G. Gudbergsson, pers.
comm.). However, other Icelandic lakes of very small
size have been found to have two different morphs of
Salv. alpinus (Galtaból = 1.4 km2, Gíslason et al.,
1999; Vatnshlídarvatn = 0.7 km2, Jónsson & Skúlason, 2000). Thus, it is possible for discrete phenotypes
to evolve and maintain genetic distinction in lakes of a
very limited size (see also Schilewan, Tautz & Pääbo,
1994), and the small size of Frostastaðavatn should
not constrain the evolution of different morphs.
Thingvallavatn was probably isolated from external
invasions of fish as early as 9130 (± 260) years ago
when the outlet was dammed by lava (Sæmundsson,
1992). The age and history of Frostastadavatn is far
less certain. What is known, though, is that the lava
habitat in the lake was formed after lava flowed into
the lake in the year 1477 AD. This means that the
scope for habitat-specific divergence of the
G. aculeatus in Frostastaðavatn is at the most about
250 generations, i.e. at least ten times shorter than it
is in Thingvallavatn. Evolutionary responses to
changes in predatory threat can certainly be significant in such a short time (e.g. Magurran, 1999;
Kristjánsson, Skúlason & Noakes, 2002b). This, and
the fact that habitat morphs in Icelandic G. aculeatus
seem to be related (Kristjánsson, 2001; Kristjánsson
et al., 2002a), suggest that we should be observing
more distinct habitat morphs in Frostastaðavatn. The
fact that we do not may stem from the lack of predation threat.
The origin of the introduced piscine predators in
Frostastaðavatn is unknown; therefore we have no
information on the level of experience of G. aculeatus
in piscivorous predation. However, as mentioned earlier, stomach content analysis of Salv. alpinus in
Frostastaðavatn from 1994 showed that only 4% of
their diet consisted of G. aculeatus (B. Jónsson,
unpubl. data). In contrast, up to 99% (AFDW) of the
diet of the large limnetic morph of Salv. alpinus in
Thingvallavatn was composed of G. aculeatus (Malmquist et al., 1992). We can reasonably conclude that
the risk of predation is lower in Frostastadavatn than
it is in Thingvallavatn. The difference in the degree of
antipredator behaviour between G. aculeatus in these
two lakes clearly supports this conclusion.
A small fish, such as G. aculeatus, that is exposed to
intense predation from piscivorous fish species will be
driven to seek refugia and food in sheltered habitats
(Werner et al., 1983; Sandlund et al., 1992). In lakes
where more than one protective habitat is offered, the
apparent competition to seek shelter can quickly promote local adaptation to the respective resources
(Holt, 1977). In Thingvallavatn, there is a great
expanse of open water, densely populated by piscivorous Salv. alpinus and some Salmo trutta, separating
the protective vegetated Nitella stands and the littoral
lava habitat. G. aculeatus have not been found in this
open pelagic zone of Thingvallavatn (Sandlund et al.,
1992; B. K. Kristjánsson & L. I. Doucette, unpubl.
data). The importance of macrophyte vegetation in
reducing predation on small fish has been shown
(Werner et al., 1983), and there is likely competition
for the sheltered habitats offered by both the Nitella
and lava habitats and avoidance of the open water.
Jakobsen, Johnsen & Larsson (1987) found that introduction of salmonid predators induced a habitat shift
in G. aculeatus from open to sheltered areas in a small
lake in western Norway. Thus, the high risk of predation in Thingvallavatn could be preventing the
G. aculeatus occupying the two protective habitats
from intermixing and interbreeding, thereby accelerating the divergence process. In Frostastaðavatn, the
overall risk of predation appears to be lower and the
two protective habitats are in closer proximity to each
other. G. aculeatus occupying the two habitats may be
able to move between them with a lower risk of predation. The result may be a lower degree of isolation
between the two groups. This may reduce the degree of
specialization of the Frostastaðavatn G. aculeatus for
their respective habitats and may greatly reduce the
likelihood, or at least the rate of, sympatric diver-
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
DIVERGENCE OF STICKLEBACKS
gence. Perhaps the morphs in Thingvallavatn and
Frostastaðavatn are at ‘different stages of segregation’, with the segregation more advanced in
Thingvallavatn.
In view of the above ecological considerations and
our behavioural data, we emphasize that a high risk of
predation in Thingvallavatn may have played a major
role in the evolution of unique and discrete foraging
behaviour and morphology in this lake, as the two
forms specialized for two different habitats with
reduced predation risk. Conversely, the lack of predation risk, and the closer proximity of protective habitats to each other, may have contributed to the low
divergence seen in Frostastaðavatn.
We conclude that risk of predation may play an
important role in sympatric divergence of fish, especially for species in which adult body size is small and
these smaller prey fish are forced to seek refugia and
food in sheltered habitats. The combined effects of ecological factors and predation pressure on morph formation and sympatric species divergence have very
likely been underestimated in the past. It would be
interesting to re-examine some studies on resource
polymorphisms and species divergence in species of
small fish, such as cichlids (Meyer, 1993), considering
the potential impact of predation risk and the relative
locations of alternative habitats on the evolution of
polymorphisms and speciation.
ACKNOWLEDGEMENTS
Thanks to Bjarni K. Kristjánsson for his detailed comments on the manuscript and his invaluable assistance in the field. Thanks to Guðbjörg Á. Ólafsdóttir
for her help in G. aculeatus collection, Bjarni Jónsson
and Hilmar Malmquist for providing valuable information on the lakes and marine sites in this study, and
Hrefna Sigurjónsdóttir for providing training and
access to essential behaviour recording software.
Funding was provided through a Graduate Student
Research Grant from RANNIS (The Icelandic
Research Council) and by The Research Fund of the
University of Iceland.
REFERENCES
Abrams PA. 2000. Character shifts of prey species that share
predators. American Naturalist 156 (Suppl.): S45–S61.
Adalsteinsson H, Jónasson PM, Rist S. 1992. Physical
characteristics of Thingvallavatn, Iceland. Oikos 64: 121–
135.
Andraso GM. 1997. A comparison of startle response in two
morphs of the brook stickleback (Culaea inconstans):
Further evidence for a trade-off between defensive morphology and swimming ability. Evolutionary Ecology 11:
83–90.
201
Andraso GM, Barron JN. 1995. Evidence for a trade-off
between defensive morphology and startle-response performance in the brook stickleback (Culaea inconstans). Canadian Journal of Zoology 73: 1147–1153.
Barber I, Huntingford FA. 1995. The effect of Schistocephalus solidus (Cestoda: Pseudophyllidea) on the foraging and
shoaling behaviour of three-spined sticklebacks, Gasterosteus aculeatus. Behaviour 132: 1223–1239.
Barber I, Huntingford FA, Crompton DWT. 1995. The
effect of hunger and cestode parasitism on the shoaling decisions of small freshwater fish. Journal of Fish Biology 47:
524–536.
Bell MA, Foster SA. 1994a. The evolutionary biology of the
threespine stickleback. Oxford: Oxford University Press.
Bell MA, Foster SA. 1994b. Introduction to the evolutionary
biology of the threespine stickleback. In: Bell M, Foster S,
eds. The evolutionary biology of the threespine stickleback.
Oxford: Oxford University Press, 1–27.
Brown JS, Vincent TL. 1992. Organization of predator-prey
communities as an evolutionary game. Evolution 46: 1269–
1283.
Bush G. 1975. Modes of animal speciation. Annual Review of
Ecology and Systematics 6: 339–364.
Bush GL. 1994. Sympatric speciation in animals: new wine in
old bottles. Trends in Ecology and Evolution 9: 285–288.
Dobhansky T. 1940. Speciation as a stage in genetic divergence. American Naturalist 74: 312–321.
Doebeli M, Dieckmann U. 2000. Evolutionary branching
and sympatric speciation caused by different types of ecological interactions. American Naturalist 156 (Suppl.): S77–
S101.
Doucette LI. 2001. Variability in resource-based behaviour of
the threespine stickleback (Gasterosteus aculeatus) in Iceland. MSc Thesis, University of Iceland.
Endler JA. 1995. Multiple-trait coevolution and environmental gradients in guppies. Trends in Ecology and Evolution 10:
22–29.
Futuyma DJ. 1986. Evolutionary biology. Sunderland, MA:
Sinauer Associates, Inc.
Gíslason D, Ferguson M, Skúlason S, Snorrason SS. 1999.
Rapid and coupled phenotypic and genetic divergence in Icelandic Arctic charr (Salvelinus alpinus). Canadian Journal
of Fisheries and Aquatic Science 56: 2229–2234.
Holt RD. 1977. Predation, apparent competition, and the
structure of prey communities. Theoretical Population Biology 12: 197–229.
Jakobsen P, Johnsen G, Larsson P. 1987. Effects of predation risk and parasitism on the feeding ecology, habitat use,
and abundance of lacustrine threespine stickleback (Gasterosteus aculeatus). Canadian Journal of Fisheries and
Aquatic Science 45: 426–431.
Johnson PA, Gullberg U. 1998. Theory and models of sympatric speciation. In: Howard D, Berlocher S, eds. Endless
forms: species and speciation. New York: Oxford University
Press, 79–89.
Jónasson PM, Adalsteinsson H, Jónsson GS. 1992. Production and nutrient supply of phytoplankton in subarctic,
dimictic Thingvallavatn, Iceland. Oikos 64: 162–187.
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
202
L. I. DOUCETTE ET AL.
Jónsson B, Skúlason S. 2000. Polymorphic segregation in
Arctic charr Salvelinus alpinus (L.) from Vatnshlídarvatn, a
shallow Icelandic Lake. Biological Journal of the Linnean
Society 69: 55–74.
Klepacker T. 1993. Morphological changes in a marine population of threespined stickleback, Gasterosteus aculeatus,
recently isolated in freshwater. Canadian Journal of Zoology
71: 1251–1258.
Kondrashov AS, Yampolsky LY, Shabalina SA. 1998. On
the sympatric origin of species by means of natural selection.
In: Howard D, Berlocher S, eds. Endless forms: species and
speciation. New York: Oxford University Press, 90–98.
Koopman K. 1949. Natural selection for reproductive isolation between Drosophila pseudoobscura and Drosophila persimilis. Evolution 4: 135–148.
Koopmans LH. 1987. Introduction to contemporary statistical
methods. Boston: Duxbury Press.
Kristjánsson BK. 2001. Divergence of Icelandic threespine
stickleback, Gasterosteus aculeatus, to two substrate types in
lakes and recently formed lagoon. MSc Thesis, University of
Guelph.
Kristjánsson BK, Skúlason S, Noakes DLG. 2002a. Morphological segregation of Icelandic threespine stickleback
(Gasterosteus aculeatus L.). Biological Journal of the Linnean Society 76: 247–257.
Kristjánsson BK, Skúlason S, Noakes DLG. 2002b. Rapid
divergence in a recently isolated population of threespine
stickleback (Gasterosteus aculeatus L.). Evolutionary Ecology Research 4: 659–672.
Lehner P. 1996. Handbook of ethological methods. Cambridge: Cambridge University Press.
Magurran AE. 1990. The adaptive significance of schooling as
an anti-predator defence in fish. Annuals Zoologica Fennici
27: 51–66.
Magurran AE. 1999. The causes and consequences of geographic variation in antipredator behavior. In: Foster S,
Endler J, eds. Geographic variation in behavior. New York:
Oxford University Press, 139–163.
Magurran AE, Pitcher TJ. 1987. Provenance, shoal size and
the sociobiology of predator evasion behaviour in minnow
shoals. Proceedings of the Royal Society of London, Series B
229: 439–465.
Magurran AE, Seghers BH. 1994. Predator inspection
behaviour covaries with schooling tendency amongst the
wild guppy, Poecilia reticulata, populations in Trinidad.
Behaviour 128: 121–134.
Malmquist HJ, Snorrason SS, Skulason S, Jonsson B,
Sandlund OT, Jonasson PM. 1992. Diet differentiation
in polymorphic arctic charr, Salvelinus alpinus, in
Thingvallavatn, Iceland. Journal of Animal Ecology 61:
21–35.
Maynard Smith J. 1966. Sympatric speciation. American
Naturalist 100: 637–650.
Mayr E. 1963. Animal species and evolution. Cambridge, MA:
Harvard University Press.
McPhail JD. 1969. Predation and the evolution of a stickleback (Gasterosteus). Journal of the Fisheries Research Board
of Canada 26: 3183–3208.
McRobert S, Bradner J. 1998. The influence of body coloration on shoaling preferences in fish. Animal Behaviour 56:
611–615.
Meyer A. 1993. Phylogenetic relationships and evolutionary
processes in East African cichlid fishes. Trends in Ecology
and Evolution 8: 279–284.
Peeke HVS, Morgan LE. 2000. Behavioural differentiation of
adjacent marine and fluvial populations of threespine sticklebacks in California: a laboratory study. Behaviour 137:
1011–1027.
Pitcher T. 1992. Who dares, wins: the function and evolution
of predator inspection behaviour in shoaling fish. Netherlands Journal of Zoology 42: 371–391.
Pitcher TJ, Parrish JK. 1993. Functions of shoaling behaviour in teleosts. In: Pitcher TJ, ed. Behaviour of teleost fishes,
2nd edn. London: Chapman & Hall, 363–439.
Ranta E, Juvonen S-K, Peuhkuri N. 1992a. Further evidence for size-assortative schooling in sticklebacks. Journal
of Fish Biology 41: 627–630.
Ranta E, Lindstrom K, Peuhkuri N. 1992b. Size matters
when three-spined sticklebacks go to school. Animal Behaviour 43: 160–162.
Reimchen TE. 1983. Structural relationships between spines
and lateral plates in threespine stickleback (Gasterosteus
aculeatus). Evolution 37: 931–946.
Rosenzweig M. 1987. Habitat selection as a source of biological diversity. Evolutionary Ecology 1: 315–330.
Sæmundsson K. 1992. Geology of the Thingvallavatn area.
Oikos 64: 40–68.
Sandlund OT, Jonasson PM, Jonsson B, Malmquist HJ,
Skúlason S, Snorrason SS. 1992. Threespine stickleback
Gasterosteus aculeatus in Thingvallavatn: habitat and food
in a lake dominated by arctic charr Salvelinus alpinus. Oikos
64: 365–370.
Sandlund OT, Jonsson B, Malmquist HJ, Gydemo R, Lindem T, Skúlason S, Snorrason SS, Jónasson PM. 1987.
Habitat use of arctic charr Salvelinus alpinus in Thingvallavatn, Iceland. Environmental Biology of Fishes 20: 263–
274.
Schilewan U, Tautz D, Pääbo S. 1994. Sympatric speciation
suggested by monophyly of crater lake cichlids. Nature 368:
629–632.
Schluter D. 1998. Ecological causes of speciation. In: Howard
D, Berlocher S, eds. Endless forms: species and speciation.
New York: Oxford University Press, 114–129.
Seghers BH. 1974. Schooling behaviour in the guppy (Poecilia
reticulata): an evolutionary response to predation. Evolution
28: 486–489.
Skúlason S, Antonsson T, Gudbergsson G, Malmquist
HG, Snorrason SS. 1992. Variability in Icelandic arctic
charr. Icelandic Agriculture Science 6: 143–153.
Skúlason S, Snorrason SS, Jónsson B. 1999. Sympatric
morphs, populations, and speciation in freshwater fish with
emphasis on arctic charr. In: Magurran AE, May RM, eds.
Evolution of biological diversity: Oxford: Oxford University
Press, 70–92.
Skúlason S, Snorrason SS, Ota D, Noakes DLG. 1993.
Genetically based differences in foraging behaviour among
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203
DIVERGENCE OF STICKLEBACKS
sympatric morphs of arctic charr (Pisces: Salmonidae). Animal Behaviour 45: 1179–1192.
Smith T. 1993. Disruptive selection and the genetic basis of
bill size polymorphism in the African finch Pyrenestes.
Nature 363: 618–620.
Snorrason SS, Skúlason S. 2004. Adaptive speciation in
northern fresh water fishes – patterns and processes. In:
Dieckmann U, Doebeli M, Metz JAJ, Tautz D, eds. Adaptive
speciation. Cambridge: Cambridge University Press, in
press.
Snorrason SS, Skulason S, Jonsson B, Malmquist H,
Jonasson P, Sandlund OT, Lindem T. 1994. Trophic specialization in Arctic Charr Salvelinus alpinus (Pisces; Salmonidae): morphological divergence and ontogenetic niche
shifts. Biological Journal of the Linnean Society 52: 1–18.
Sokal RR, Rohlf FJ. 1994. Biometry: the principles and
203
practice of statistics in biological research. New York:
W.H. Freeman.
Taylor EB, McPhail JD. 1986. Prolonged and burst swimming in anadromous and freshwater threespine stickleback,
Gasterosteus aculeatus. Canadian Journal of Zoology 64:
416–420.
Walker JA. 1997. Ecological morphology of lacustrine
threespine stickleback Gasterosteus aculeatus L. (Gasterosteidae) body shape. Biological Journal of the Linnean Society 61: 3–50.
Ward A, Krause J. 2001. Body length assortative shoaling in
the European minnow, Phoxinus phoxinus. Animal Behaviour 62: 617–621.
Werner E, Gilliam J, Hall D, Mittelbach G. 1983. An experimental test of the effects of predation risk on habitat use in
fish. Ecology 64: 1540–1548.
© 2004 The Linnean Society of London, Biological Journal of the Linnean Society, 2004, 82, 189–203