Habitat use and feeding behaviour in two closely related fish species, the three-spined and nine-spined stickleback: an experimental analysis


Paul J. B. Hart, Department of Biology, University of Leicester, Leicester LE1 7RH, UK. Tel. +44 01162 523348; Fax: +44 01162 523330; E-mail: pbh@le.ac.uk


  • 1In this paper I analyse experimentally the way in which nine-spined sticklebacks (Pungitius pungitius) and three-spined sticklebacks (Gasterosteus aculeatus) use weeded areas and open water and how this use is influenced by food availability.
  • 2Experiments were designed to answer two questions: what preference do the two stickleback species have for weeded areas and the open water and then how is this preference changed by the presence or absence of food? Further, differences in behaviour between the two species were analysed.
  • 3Results showed differential habitat use by the two species: nine-spined occupied the weeded area more often than three-spined and were less likely to utilize the bottom 10 cm of the water column.
  • 4While foraging, nine-spined swam and pointed more often than did three-spined and modified some of these behaviours in response to the distribution of food.
  • 5The nine-spined stickleback has a slimmer, more streamlined body than does the three-spined and the differences in habitat use and behaviour reflected this.
  • 6The results are discussed in relation to interspecific competition and speciation through the ecological selection of different trophic morphologies.


Understanding competition is of fundamental importance to ecology as it is an important factor determining the number of species that can be packed into a habitat, so determining the upper limits to biodiversity (Werner 1977; Pimm 1991; Persson 2002). When two species appear to live in the same habitat and eat almost the same diet, the immediate reaction is to search for characteristics that separate the ecologies of the two. Gause's principle states that two species cannot have exactly the same niche (MacArthur & Wilson 1967; MacArthur 1972; Schluter 2000). Character displacement can push apart the phenotypes of two closely related species that live together and compete for similar resources (Schluter & McPhail 1992; Schluter 2000). Darwin's finches are the most famous example of this phenomenon (Grant 1986). Disentangling the characteristics of two closely related sympatric species can be a complex task, as differences can be subtle, multivariate and can vary in their effect though time.

In this paper I analyse experimentally the way in which nine-spined sticklebacks (Pungitius pungitius Linnaeus 1758) and three-spined sticklebacks (Gasterosteus aculeatus Linnaeus 1758) use two types of habitat and how habitat choice is influenced by food availability. The two species are often found together in both streams and still water (Hynes 1950; Walsh & FitzGerald 1984; Delbeek & Williams 1987; Copp et al. 1998; Copp & Kováč 2003; Chambers unpublished data). Detailed studies of the diet of the two species show that they eat almost the same range of foods (e.g. Hynes 1950) but they show differences in the way in which they use the habitat (Delbeek & Williams 1987; Copp et al. 1998; Copp & Kováč 2003). Nine-spined sticklebacks have a greater preference for weeded areas, whereas the three-spined stickleback is found more often in open water. There is a correlation between the differential use of cover and bodily defence against predators. Both species have spines on the dorsal surface and on the pelvic girdle; however, these are longer in the three-spined stickleback (Maitland & Campbell 1992; Bowne et al. 1994). The functional significance of these differences was demonstrated first in experiments by (Hoogland, Morris & Tinbergen 1957), who showed that nine-spined sticklebacks were much more vulnerable to predation by pike (Esox lucius) and perch (Perca fluviatilus) than were three-spined sticklebacks.

The difference in armour is not the only morphological difference between the two species. Nine-spined sticklebacks are generally smaller in length, are shallower dorso-ventrally and have a slimmer caudal peduncle (Maitland & Campbell 1992). These differences probably mean that the nine-spined stickleback has different swimming capabilities than does the three-spined, whose deeper body facilitates rotation around the vertical axis to pick items off the bottom of a lake. Similar differences are found between the benthic and limnetic morph of the three-spined stickleback in lakes in British Columbia, Canada (Schluter & McPhail 1992). The limnetic species has a smaller, slimmer body and is best at capturing planktonic prey, while the benthic type is most efficient at catching larger benthic dwelling prey (Schluter & McPhail 1992; Schluter 1993). In the case of the nine-spined and three-spined sticklebacks, the parallel with the limnetic and benthic morphs of the Canadian three-spined is not exact, as the nine-spined stickleback prefers sheltered habitats while the morphologically similar limnetic three-spined is a denizen of open waters.

According to some taxonomists the nine- and three-spined sticklebacks are sibling species which might explain the closeness of their ecologies (see Bowne 1994 for a review), but McLennan & Mattern (2001) combine behavioural and morphological characters to produce a phylogeny that puts Pungitius with the genus Culea and G. aculeatus with G. wheatlandi. Hybrids have been produced between nine- and three-spined sticklebacks but the offspring, which do grow to maturity, rarely reproduce themselves.

The experiment described in this paper was designed to answer two questions. First, I wanted to examine the habitat preferences of the two species in the context of presence or absence of cover and secondly, how this preference would be changed by the presence or absence of food in the two habitats. In answering both questions I wanted to look at the preferences shown by the fish when no potential competitor was present. In other words, the experiments have attempted to show what basic preferences the two species bring with them when they come into contact in the natural environment. Any interspecific interaction will change the basic preference, as the other species becomes a part of the subject's environment. Understanding interactions will be helped by knowledge of the basic preferences with which each competitor starts.



The experiments used two 600 L aquaria housed indoors with artifical lighting from two 60 W bulbs suspended 30 cm above each tank. It was light in the room from 06.00 to 21.00 hours each day. Water temperature was 20·3 °C (± SD 0·5) for the three-spined trials and 22·4 °C (± SD 1·0) for the nine-spined trials. Each tank was 2 m (l) × 0·48 m (h) × 0·6 m (b) and had a grid made of white tape on one long side. The grid was divided into 4 (down) × 11 (along) squares, the majority of which were 10 × 20 cm. The remaining three sides of the tank were covered on the outside by black plastic sheets. The bottom of the tank was covered with medium gravel and in the centre of the tank a rectangular tray (0·455 m (l) × 0·225 m (b)) filled with more gravel held five artificial plastic aquatic plants which floated so as to provide a weeded area extending from the bottom to the surface of the water. The weeded area occupied divisions 5–7 as seen from the grid on the front of the tank and had open water all around it.

experimental animals

Three-spined sticklebacks were caught with bottle traps from a pond outside the Science Library (UB2) of the University of Lund, Sweden. Nine-spined sticklebacks were also caught with bottle traps but from ponds in the University's Botanic Garden. All fish were kept in the laboratory for at least 2 weeks before the experiments began and were introduced to the experimental procedure at least twice before data was gathered from them. The mean length of the three-spined sticklebacks was 58·8 mm (± SD 1·8) and for the nine-spined sticklebacks 47·9 mm (± SD 3·26).

Benthic prey, collected from the Lund Water Works storage tanks, was represented by Asellus aquaticus, and individuals used were between 2 and 6 mm long. 2–3 mm long Daphnia spp., also collected from the Water Works were used as the limnetic prey.

experimental design

The experiment was designed to reveal the habitat preferences of the two species and how they were influenced by the presence or absence of food. The combinations of prey used and their location in the tank is shown in Table 1. The four treatments were each replicated eight times using the same eight fish. These were selected at random for each replicate, so individuals were never used in the same order. All the replicates for one treatment were completed before moving on to the next treatment. Experiments started with Treatment D, then B, then C and finally A. Tank availability prevented treatment randomization. Removing prey not required in the next trial from a tank was time-consuming and it was difficult to guarantee that all uneaten prey had been found. The schedule of trials chosen minimized the likelihood that stray prey from previous trials were present. All the trials with three-spined stickleback were done first followed by the trials with nine-spined.

Table 1.  The experimental design used to test habitat choice by three-spined and nine-spined sticklebacks and the influence on this choice of the presence or absence of food. The numbers are the number of each prey type with D = Daphnia (open water only) and A = Asellus (weed only)
Prey in open waterPrey in weed
 600D/100 A600D/0 A
 0D/100 A0D/0 A

experimental protocol

At the start of a series of trials Asellus and Daphnia were placed in the weed or open water in the combinations required (see Table 1). Prey was left to distribute for 30 min. After each replicate, prey numbers were augmented by the number of prey eaten during the 20-min trial. At the start of a trial a subject fish was caught from the holding tank with a hand net and transferred to the experimental aquarium. The fish was not released immediately but was transferred first to a small cage held at the water surface. After 5 min the cage was tilted and the fish allowed to swim out into the aquarium.

Data was recorded onto a sheet with 40 rows. Each row was used to record what the fish was doing at the end of each half-minute interval (instantaneous or point sampling; Lehner 1996). The behaviours recorded are shown in Table 2. At the same time the position of the fish relative to the grid on the front of the tank was recorded in the form of a letter from A to D and a number from 1 to 11 where A was the bottom row and 1 the left hand column. A record was also made of whether or not the fish was in the weed.

Table 2.  The behaviours recorded every 30 s during the 20-min observation period
SwimMoving through the water using either tail thrusts or propulsion from the pelvic fins
PointStationary and positioned with the˙ fish focusing on some object
HangStationary but with the long axis of the body horizontal to the substrate. Fish not focusing on any object
EatIngest a prey item

At the end of the 20-min session the fish was removed from the aquarium and placed in a holding tank where it remained until all eight replicates had been completed. The fish were then transferred back to their home tank where they were fed to satiation. This ensured a constant internal state at the start of each set of trials.

data analysis

The data were analysed using a general linear model (GLM, Grafen & Hails 2002) in which the response variables were the number of sampling instants, of a total of 40, at which a behaviour had been displayed. The independent variables were species and the presence or absence of the two prey types. All data were transformed to ln(x + 1) to deal with non-normality.

For the purposes of analysis I treated the number of occurrences of a state as independent, reasoning that the probability of each of the states being expressed at the instant of observation was equal to the proportion of time the animal spent in a particular state. The gap between samples was long enough to allow the fish to change state several times between one sampling point and the next. Consequently, being in a certain state at sampling instant n would not have determined the state recorded at instant n+ 1.


There were two null hypotheses: (1) that habitat choice by the two species did not differ from each other and (2) that the availability of either Asellus or Daphnia or both did not influence habitat choice in either species. The best measure of habitat choice recorded was the number of sampling instants at which the subject fish was in the weed patch and the results are shown in Fig. 1a. Analysis showed that there were significant differences between the species (FS = 14·64, P = 0·000, d.f. = 1). The presence of Asellus introduced significant variation (FA = 7·83, P = 0·007, d.f. = 1) but the presence of Daphnia had no effect (FD = 2·35, P = 0·131, d.f. = 1). The interactions between species and both prey types were not significant and this was true for all subsequent analyses with P > 0·05 in all cases.

Figure 1.

Habitat use by nine- and three-spined sticklebacks. (a) The median number of 30-s observations when either nine-spined (white bars) or three-spined sticklebacks (grey bars) were in the weed patch. (b) The median number of 30-s observations when either nine-spined or three-spined sticklebacks were within 10 cm of the tank bottom. The vertical lines show the 1st and 3rd quartile on the medians.

As Fig. 1a shows, nine-spined sticklebacks were more often in the weed. Even when there was no food in the tank, the nine-spined spent time in the weed. In comparison, three-spined sticklebacks went into the weed only when it contained prey. The lowest occupancy of the weed by nine-spined was when only limnetic prey were available in the open water.

Another aspect of habitat use examined was where in the water column the two species were most often found. The number of sampling instants that the fish were in sector A of the tank, representing the bottom 10 cm of the water column, was significantly different between the two species (Fig. 1b; FS = 19·8, P = 0·000, d.f. = 1). Neither prey type had a significant effect (FA = 0·46, P = 0·5, d.f. = 1; FD = 0·15, P = 0·696, d.f. = 1). This result showed that three-spined sticklebacks were recorded significantly more often in the bottom 10 cm of the water column than were the nine-spined.

A secondary null hypothesis was that the two species showed no differences in behaviour and that the fish did not change their behaviour in response to prey type and availability. I now look at the differences between the species in the behaviours recorded.

The GLM for eating showed that there was no difference in the eating pattern between the species (FS = 1·62, P = 0·28, d.f. = 1; Fig. 2a). The presence of food did have a significant influence (FA = 11·43, P = 0·001, d.f. = 1; FD = 46·1, P = 0·000, d.f. = 1). Feeding rate was highest when both prey types were available, next highest with just Daphnia, then with just Asellus and finally there was a residual feeding activity when there were no prey in the aquarium. In this last case most of the activities recorded as feeding were fish biting at floating inorganic particles or at bubbles.

Figure 2.

Differences in feeding behaviour between nine- and three-spined sticklebacks. (a) The median number of 30-s observations when either nine-spined (white bars) or three-spined sticklebacks (grey bars) were eating. (b) The median number of 30-s observations when either nine-spined or three-spined sticklebacks were swimming. (c) The median number of 30-s observations when either nine-spined or three-spined sticklebacks were pointing. The vertical lines show the 1st and 3rd quartile on the median for each behaviour.

There were significant differences in swimming behaviour between species (Fig. 2b; FS = 20·22, P = 0·000, d.f. = 1). The presence of Asellus and Daphnia also had significant effects (FA = 6·83, P = 0·011, d.f. = 1; FD = 12·81, P = 0·001, d.f. = 1). Three-spined sticklebacks swam less than did nine-spined and swimming occurred less often when there was limnetic food available.

Pointing and hanging were inversely correlated and this relationship was strongly significant when one outlier value was excluded (r = 0·94, d.f. = 5, P < 0·01). As a consequence, only pointing was analysed using a GLM. There was a significant effect of species on the number of times pointing was observed (Fig. 2c; FS = 9·84, P = 0·003, d.f. = 1). Neither prey type had an effect on this behaviour (FA = 2·77, P = 0·102, d.f. = 1; FD = 2·02, P = 0·16, d.f. = 1). Three-spined pointed less than did nine-spined, which means that three-spine were more often observed hanging.


The experiments showed that the two species of stickleback have significant differences in habitat use and foraging behaviour. Nine-spined sticklebacks used the weed even when there was no food present (Fig. 1a), while the three-spined stickleback went into the weed only when it contained food. Three-spined tended to keep within 10 cm of the bottom for most of the time while nine-spined sticklebacks were more often higher in the water column (Fig. 1b). While foraging, nine-spined swam more often and pointed more often than did three-spined (Fig. 2) and modified some of these behaviours in response to the distribution of food. The nine-spined stickleback has a slimmer, more streamlined body than does the three-spined and the differences in habitat use and behaviour reflected this.

It should be noted that the treatments were not randomized, meaning that order and treatment are correlated. As each individual of both species experienced the same treatment order, differences between species should be unaffected. The three-spined fish hardly visited the weed in treatments D and B in which no prey was present (Fig. 1a). Only when Asellus was in the weed in treatments A and C did the three-spined fish venture into it. If order were important, overriding the treatment effect, then I would have expected the three-spined sticklebacks to make fewer visits into the weed in treatment C than in treatment D having just experienced two trials with no food in the weed. In fact, the reverse is true, which implies that the fish are responding to the conditions of the treatment rather than to their experience. Similarly, the nine-spined fish behaved in the treatments as one would expect if they were responding to the treatment conditions rather than to their experience. The smallest number of times they were observed in the weed was for the treatment when the only prey was in the open water (B). In treatment C, with only Asellus in the weed, the amount of time the nine-spined fish spend in the weed increased to above the level recorded in treatment D when no prey were present. Again, this supports the interpretation that fish are responding to treatment and not to experience or order.

The sequence of behaviours used by three-spined sticklebacks, derived from Gill & Hart (1994), is shown in Fig. 3. The bold-edged boxes are the behaviours recorded in this study. Personal observations show that the nine-spined stickleback shows all the behaviours employed by the three-spined. The figure illustrates that the feeding behaviours used by the two species is different only in the time allocated to behaviours. Figure 2 shows that the presence of food drives most of the differences observed in the cycle. Naturally enough, fish fed more when food was available, but were more often caught feeding when the number of prey items was high. This was the case when Daphnia was in the limnetic zone (Fig. 2). It is also possible that when only the relatively large Asellus were available, the fish satisfied their hunger more quickly. It has been shown by Hart & Ison (1991), Hart & Gill (1992, 1993) and Salvanes & Hart (1998) that as the stomach of a three-spined stickleback fills, the rate of feeding falls and the fish chooses only smaller prey items.

Figure 3.

The feeding cycle proposed by Gill & Hart (1994) for three-spined sticklebacks modified to show the differences in behaviours between nine-spined and three-spined sticklebacks. The behaviours with bold frames are those examined in this study. Those with dotted frames are the behaviours shown to be involved in the more detailed study of sequences. Where two arrows or lines emerge from a box it signifies that a decision is made at this point which has two outcomes shown by the behaviour at the end of the respective arrows.

The most important result for interpreting the relationship of the two species in the wild is their different use of space. The nine-spined sticklebacks position themselves more often in the weeds and in the upper part of the water column, while the three-spined stays close to the bottom and in the open. In the wild the two species will be exploiting different sets of prey even though those sets might contain the same items. When food abundance is high, the overlap in the resource base exploited by the two species will not be significant. Recent evidence gathered by Chambers (unpublished results) shows that in some instances of three- and nine-spined sympatry, there is evidence of character displacement in three-spined sticklebacks in those morphological characters that are important for foraging. This suggests that competition between the two species leads to the three-spined stickleback becoming more limnetic in its feeding habits and developing a narrower jaw and more gill rakers.

The experiments have shown that nine-spined and three-spined sticklebacks differ in how they use the habitat. This, coupled with differences in reproductive behaviour (Ketele & Verheyen 1985; Maitland & Campbell 1992) and poor fertility of hybrids, is probably sufficient to keep the two species apart. The species pair studied mirrors the way in which the G. aculeatus group has been split into forms, which differ in trophic morphology, reproductive behaviour and habitat use (McKinnon & Rundle 2002). The nine-spined and three-spined differ from their counterparts within the Gasterosteus genus by being separated by a longer period of evolutionary time. Despite this, the differences between G. aculeatus and P. pungitius are of the same type as found in the less permanent forms existing within the three-spined stickleback group.

The results of this study show that the populations of the two species used in the experiments have habitat and resource uses that overlap. If the two populations are representative of the general stickleback case then my results show that the two species are found to coexist over the long term in nature despite the demonstrated habitat and resource use overlap. That the two species do compete for food is shown by the existence of character displacement in the three-spined stickleback when together with the nine-spined. Character displacement shows that the niche occupied by a species is determined to some degree by the characteristics of the other species with which it finds itself (Pimm 1991). Species packing in a community is not a predetermined characteristic independent of the organisms in it, but one dependent on the interactions between the characteristics of the species making up the community.


I would like to thank Nichola Mandle for assistance with the experiments and Stellan Hamrin for providing accommodation at the Limnology Section, Department of Ecology, University of Lund, Sweden. The visit to Lund was funded by a grant from the Royal Society of London. I am grateful to Lennart Persson, Helen Chambers and Ashley Ward for reading earlier versions of the manuscript and to Robert Smith and André Gilburn for advice on statistics.