Introduction

Shoaling behaviour provides benefits to the individual shoal members: in surface-dwelling fish, shoaling often serves to avoid predation (Brock and Riffenburgh 1960; Neill and Cullen 1973; Seghers 1974; Magurran et al. 1985; Godin 1986; Magurran and Pitcher 1987; Magurran 1990; Pitcher and Parrish 1993; Krause and Ruxton 2002). For example, in the guppy (Poecilia reticulata) the tendency to shoal is less pronounced in habitats with low predation risk (Seghers 1974; Magurran and Seghers 1991; Magurran et al. 1992, 1993). Furthermore, the ability to find food (Keenleyside 1979; Pitcher et al. 1982; Pitcher and Parrish 1993) or the ability to occupy a food source guarded by a physically stronger competitor (Foster 1985; Chapman and Kramer 1996) may increase with increasing shoal size. Additionally, shoaling fish can learn about both predators and prey from conspecifics (e.g., Brown and Laland 2003 for a review).

However, costs of group living arise from competition for food between the individual members of a shoal (see Pulliam and Caraco 1984; Ranta et al. 1993; Krause and Ruxton 2002). One would predict that shoaling behaviour should be reduced if the costs of increased food competition overrun the benefits of lower predation risk and increased food-finding ability. In fact, starved fish tend to shoal less than fish in a good nutritional state, probably because they are more affected by competition from shoal members (van Havre and FitzGerald 1988; Krause 1993; Reebs and Saulnier 1997; Krause et al. 1999).

The study system: habitat differences

Cave habitats differ from many surface habitats in the lack of photoautotrophic primary production due to the absence of light. In most cave ecosystems organic matter washed in from the surface (like leaves) and bat guano are the trophic basis of the subterranean food web (Poulson and White 1969; Parzefall 1993a; Poulson and Lavoie 2000). As a consequence, food is scarce in most caves (Barr and Holsinger 1985). Cave fishes may evolve physiological strategies to cope with this problem, such as reduced metabolic rate (Hüppop 1986) or the development of additional fat-storing tissues (Hüppop 1988, 2000). Most importantly, small fish may suffer intense predation pressure in surface habitats (e.g. by predatory fishes and birds), but predation on cave fishes is mostly absent (Parzefall 1993a; but see Plath et al. 2003a for predation on cave-dwelling Poecilia mexicana). Consequently, benefits of shoaling behaviour in terms of reduced predation risk are lacking in most cave fishes. Furthermore, the costs of food competition can increase in cave habitats due to the low food availability. Altogether, shoaling behaviour has no obvious adaptive significance in cave fishes and should be reduced. If costs of group living are high, cave fishes might even actively avoid conspecifics. In the present article, we tested this prediction in a species of barb (Garra barreimiae Fowler and Steinitz 1956) that has evolved a cave-dwelling population. We compared the shoaling tendency between the surface-dwelling and cave-dwelling forms.

The study species

Surface-dwelling G. barreimiae are endemic to the Arabian Emirates and to the Oman (Fowler and Steinitz 1956; Feulner 1998). G. barreimiae is one of the most common species of fish in the “Widyan”, that is, river systems that periodically fall dry (Krupp and Schneider 1988). This fish has the remarkable ability to move between pools across land in search of other pools (Feulner 1998). G. barreimiae mainly feeds on algae and other epibionts (Büttiker and Krupp 1989). Algae grow in large quantities in the shallow parts of the Widyan, which are intensely penetrated by sunlight (Büttiker and Krupp 1989).

Cave fishes evolve when a previously surface-dwelling species starts to colonise the groundwater of a cave. In our study system, this means that surface-dwelling barbs probably represent the most recent ancestor of the cave form, and cave-dwelling barbs represent the derived form. However, molecular phylogenetic studies on G. barreimiae have so far not been carried out. The cave form of G. barreimiae was discovered by A. Dunsire and M. Gallagher in 1980 in a cave in the Jabal-Akhdar mountain system near Al-Hamra, Oman (Banister 1984, 1987). G. barreimiae is the only known species of fish from this cave. Predation on cave-dwelling G. barreimiae is unknown. Food availability is low in the cave habitat of G. barreimiae (Weissenbacher et al. 2002). We assume that cave-dwelling G. barreimiae feed on bat guano, organic matter washed into the cave, and small invertebrates. However, thorough studies on the ecology of the cave population are lacking.

Sensory basis for shoaling behaviour

During the colonisation of subterranean habitats, formerly surface-dwelling animals have to cope with one predominant problem: due to the absence of light all behaviour patterns that rely on vision become obsolete (Parzefall 1993a). A similar, but less dramatic scenario might occur in shoaling species that are active at night, or in species living in murky water. To compensate for the lack of visual information, other sensory systems (e.g., sense of touch, sense of smell) are often elaborated: for example, sensitive body appendices (i.e. barbells or fins in fishes) may be enlarged in cave dwellers (Weber et al. 1998; Weber 2000). The visual system becomes useless in the dark and is often reduced (e.g. Wilkens 1988; Langecker 2000). If cave fishes exhibited shoaling, they would need to use non-visual senses to detect shoal members. In aquatic cave animals, chemical or mechanical (water displacement) signals appear suitable for this purpose, but non-visual communication has not yet been examined in surface- and cave-dwelling G. barreimiae. Eyes are reduced in the cave form (Banister 1984). However, we detected small externally visible remnants of eyes in the fish reared in our laboratory. It is yet unknown whether these cave fish still perceive visual cues when raised in light. In the present article, we address the following questions on shoaling behaviour in G. barreimiae:

  1. (1)

    Does the surface form associate with a shoal of conspecifics and is this behaviour reduced in the cave form?

  2. (2)

    What types of cues (visual and non-visual) are used by the two forms?

To answer these questions, we carried out simultaneous choice tests, giving focal fish a choice of a stimulus shoal and no shoal. We generally expected reduced shoaling or even avoidance of conspecifics in the cave form. Three different treatments were performed:

  1. (1)

    Only visual cues from the stimulus shoal could be perceived.

  2. (2)

    Both visual and non-visual cues from the stimulus shoal could be perceived. Here, we wanted to assess the strength of the shoaling tendency in the surface and the cave form when multiple cues were available.

  3. (3)

    Only non-visual cues from the stimulus shoal could be perceived. These tests were carried out in darkness but allowed for the exchange of chemical and tactile information.

Methods

Study organisms and general housing conditions

We used two populations of G. barreimiae from Oman: the surface-dwelling fish were collected in 1956 by H. W. Fowler and H. Steinitz. The cave form was caught in 1980 in the Hoti cave system (23°05′N, 57°21′E), Jabal-Akhdar mountains, approximately 130 km south-southwest of Maskat, Oman by A. Dunsire and M. Gallagher. Some specimens were transferred to the aquarium of the Zoological Institute in Hamburg for H. Wilkens in 1990. The fish used in our studies came from the third to fourth laboratory generation. Test fish from both populations (surface: 6.78±0.55 cm; cave: 5.68±0.41 cm) were mature and were approximately the same age (7 years) when they were used in the tests. Stocks were maintained in 50- to 200-l aquaria containing hiding places and plants at 25–30°C. Surface and cave fish were kept in different tanks. The bottom was covered with bright gravel. Fish were raised under an artificial 12:12 h light:dark cycle, ensuring that even the cave fish were familiar with light during our tests. Fish were fed twice a day ad libitum with flake food and Artemia naupliae, occasionally supplemented by live cladocerans and Tubifex worms. We were unable to distinguish between the sexes. During the shoaling tests, sexual behaviour was not observed. However, some fish of both populations spawned some days after the experiments. This was probably a response to the previous frequent water exchange, which stimulates spawning (H. Wilkens, personal communication).

Shoaling experiments

The test tank (120×50×40 cm, length × height × width) was filled with aged tap water of 27°C. The bottom was covered with a thin layer of bright gravel. Two vertical lines, marked on the front of the tank, divided the tank into three equal compartments: two lateral preference zones and a central neutral zone. We placed a cylinder (15 cm diameter, 30 cm height) in each of the preference zones to hold the stimulus shoal. Hence, the focal fish could swim around the shoal. We used three different combinations of cylinder materials and illumination conditions (Plath et al. 2003b).

In treatment 1, we used clear Plexiglas cylinders and the tests were carried out in light. The upper part of the cylinders was above water level. The cylinders were not closed at the bottom. However, the bottom was embedded in gravel during the tests, and chemical cues were unlikely to reach the focal fish. Hence, in this treatment, the test fish could perceive only visual cues from the stimulus shoal.

In treatment 2, we used wire-mesh cylinders instead (1 mm wire diameter, 5 mm mesh width) and the tests were also performed in light. In this experiment, the focal fish could perceive both visual and non-visual cues from the stimulus shoal.

In treatment 3, the same type of mesh cylinder was used but the tests were carried out in darkness. In this experimental situation we assumed that the test fish could perceive solely non-visual cues from the shoal.

During the tests in light, the test tank was illuminated from above by a centrally located 36 W incandescent bulb, providing approximately 90 lx on the water surface. The observer sat quietly approximately 2 m from the tank. During the tests in the absence of visible light, a 500 W infrared bulb (wavelength >800 nm) was used instead. In this case, trials were recorded by an infrared-sensitive camera, transferred to a monitor in a neighbouring room, and scored directly.

Before a test, three stimulus fish were randomly taken from the stock tanks and introduced into one cylinder. Stimulus fish came from the same population as the focal fish population. The other cylinder remained empty. We waited until the stimulus fish were swimming calmly. Then, a focal fish was gently introduced into the neutral zone. When the focal fish began to swim, we started measuring the time it spent in both preference zones for 5 min. Then, the position of the two cylinders was carefully exchanged and measurement was repeated. The focal fish remained in the test tank while we exchanged the cylinders. This procedure was performed to detect side biases. Times from both parts of a trial were added. We a priori decided to exclude trials in which the fish spent more than 80% of the choice time in only one compartment during both parts of a trial due to side biases. No trials had to be excluded due to this criterion.

All fish were tested in all three treatments. Between the tests, test fish were kept in separate tanks for 1–2 days. The order of the treatments was random. Focal fish may have served as a stimulus fish in a subsequent trial. Due to the limited number of fish available for the tests (surface form: n=16, cave form: n=11), we could not use completely different stimulus fish for each focal fish tested. However, the combination of individual fish used for the stimulus shoal was random.

Statistical analysis

To test for differences between the two populations and between the three treatments, we calculated the shoaling tendency (time in the compartment with the shoal − time in the opposite compartment). We compared shoaling tendencies using population (surface or cave) as between factor and treatment (treatments 1–3) as repeated measure for a repeated measures analysis of variance (ANOVA) using StatView 5.0.

We post hoc asked whether the focal fish generally responded to the stimulus fish under the three different experimental conditions. Therefore, we tested for differences between the time spent near the shoal and the time spent in the empty compartment during each treatment using paired t tests (SigmaStat 2.0). The latter test allowed us to examine whether the measured association times were due to random distribution of the focal fish in the test tank. Random distribution would be predicted if the focal fish did not perceive any information from the stimulus shoal. Since all data sets were analysed twice, we applied a Bonferroni adjustment as follows: α′=0.05/2=0.025.

Results

Comparison of the shoaling tendency between the populations

Population had a significant influence on shoaling tendency (repeated measures ANOVA: F1,25=128.07, P<0.0001, power=1.0). A post hoc test revealed that the surface fish spent more time near the stimulus fish than did the cave fish (Tukey test: P<0.05; Fig. 1). Treatment also significantly influenced shoaling tendency (repeated measures ANOVA: F2,50=6.11, P=0.0042, power=0.88). A post hoc test revealed that treatment 1 differed from treatments 2 and 3 (Tukey test, P<0.05 in both cases), whereas treatments 2 and 3 did not differ (P>0.05). The interaction term Population × Treatment also had a significant influence on the time spent shoaling (repeated measures ANOVA: F2,50=19.24, P<0.0001, power=1.0). The surface form showed the strongest shoaling tendency during the treatment in which the shoal was presented in Plexiglas cylinders in light (treatment 1), whereas the cave form showed no shoaling in this treatment (Fig. 1).

Fig. 1
figure 1

Shoaling tendency in surface-dwelling (open bars) and cave-dwelling (black bars) Garra barreimiae (time spent near a stimulus shoal − time in an empty control compartment; mean ± SD). A stimulus shoal comprising three conspecifics was confined to (1) clear Plexiglas cylinders in light, (2) wire-mesh cylinders in light, and (3) wire-mesh cylinders in darkness.

Full size image

General response to the stimulus shoal

In treatment 1, surface-dwelling G. barreimiae spent significantly more time in the compartment with the stimulus shoal (mean ± SD: 386.88±41.94 s) than in the compartment containing the empty Plexiglas cylinder (94.94±44.53 s; paired t test: t15=18.06, P<0.001, α′=0.025, power=1.0; Fig. 1). In treatment 2, the surface fish also spent significantly more time near the stimulus shoal (337.13±63.01 s; empty cylinder: 167.50±40.37 s; t15=7.53, P<0.001, α′=0.025, power=1.0). Also in treatment 3, the test fish significantly preferred the stimulus shoal (paired t test: t15=8.33, P<0.001, α′=0.025, power=1.0). They spent 295.44±30.51 s near the stimulus shoal and 158.88±40.46 s in the empty compartment.

Cave-dwelling G. barreimiae did not prefer either compartment in treatment 1 (near shoal: 237.82±18.54 s; near empty cylinder: 234.55±17.22 s; t10=0.37, P=0.72, α′=0.025, power=0.05). In treatment 2, the cave fish spent significantly more time near the shoal (258.27±31.72 s) than near the empty cylinder (211.64±23.66 s; t10=3.00, P=0.013, α′=0.025, power=0.73). In treatment 3, the cave fish also spent significantly more time in the compartment with the shoal (244.36±18.38 s; empty: 201.64±17.59 s; t10=4.26, P=0.002, α′=0.025, power=0.97).

Discussion

Surface population

We measured the time surface-dwelling and cave-dwelling G. barreimiae spent near a stimulus shoal or in an empty compartment in a simultaneous choice situation. In surface-dwelling G. barreimiae shoaling behaviour appears to be pronounced as they strongly preferred to associate with the stimulus shoal in all three treatments. Shoaling behaviour has also been reported for a variety of other cyprinid fishes (e.g., Magurran et al. 1985; Hager and Helfman 1991; Ashley et al. 1993; Krause 1994; Chivers et al. 1995; Krause et al. 1999; Reebs and Saulnier 1997; Barber and Wright 2001).

The preference to associate with conspecifics was most pronounced in treatment 1, where the shoal was separated from the focal fish by transparent Plexiglas; it was weaker when the shoal was isolated by a mesh grid in light (treatment 2). The visibility of the stimulus shoal was to some extent better in the Plexiglas treatment than in the wire-mesh treatment at least to the observer’s eye. The surface form readily responded to solely non-visual cues (treatment 3). These fish might also shoal at night or in dim light, when non-visual cues are more reliable to locate conspecifics. Furthermore, fish from the surface form mostly swim near the bottom in their natural habitat and often hide under stones or in crevices (Feulner 1998). Here, vision may be obscured, so that surface-dwelling fish occasionally have to rely on non-visual communication. However, previous studies have demonstrated that even sea fish dwelling in open water may school when blinded (Partridge and Pitcher 1980). In summary, surface fish use both visual and non-visual cues to shoal, but shoaling is more pronounced when visual cues are available.

Cave population

In the cave form, no response to solely visual cues was found. The visual system appears to be considerably reduced, thereby making the perception of visual signals impossible even in light-reared animals (see Banister 1984). Alternatively, the perception of visual cues might still be possible but the cave fish do not respond to these cues. Qualitative observations suggest that the cave form does not perceive visual cues such as hand movements outside the tank.

When the focal fish could perceive non-visual cues (treatments 2 and 3), fish from the cave form did shoal, but their shoaling tendency was considerably weaker than that of the surface form. Like the surface-dwelling fish in darkness (treatment 2), fish from the cave form probably use pressure-wave signals to detect the stimulus shoal via their mechanosensory lateral-line system. The lateral-line system is involved in position keeping in schooling fish (Partridge and Pitcher 1980). The lateral line is modified/enhanced in many cave fishes (Poulson 1963; Walters and Walters 1965; Parzefall 1970; Thinès and Durand 1973; Weber 2000). Cave-dwelling Astyanax fasciatus use their lateral-line system for “hydrodynamic imaging” (Hassan 1989), for detecting moving objects (like other fish or prey), and for rheotaxis (Montgomery et al. 2001 for a review). However, little is known about the morphology and physiology of the mechanosensory lateral-line system in surface-dwelling and cave-dwelling G. barreimiae. Besides mechanical (water displacement) cues, chemical cues might be used to locate shoal members. For example, the Somalian cave barb Phreatichthys andruzzii (Cyprinidae) responds to chemical cues from conspecifics (Berti et al. 1989). Further studies will be needed to investigate the relative influence of mechanical and chemical cues on the detection of conspecifics in G. barreimiae.

Reduction of shoaling behaviour in the cave form

Our results confirm the prediction that the shoaling tendency would be reduced in the cave form. Even in darkness, the surface form still exhibited more pronounced shoaling behaviour than the cave form. Since the two populations used in our tests were reared under identical conditions, the differences in the shoaling tendency between surface-dwelling and cave-dwelling G. barreimiae appear to be genetically determined. A potential problem of our study was that fish had been raised in captivity for several years and generations. Hence, we discuss the genetic basis for the observed differences with caution. A reduction of the tendency to form social aggregations has also been reported for other cave fishes, such as the Mexican cave tetra Astyanax fasciatus (Characidae) and the Mexican cave molly Poecilia mexicana (Poeciliidae; Parzefall 1993b), and for the African cave fishes Caecobarbus geertsii, Barbopsis devecchii (Cyprinidae), and Uegitglanis zammaranoi (Clariidae; Jankowska and Thinès 1982). We assume that low predation risk and extremely low food availability (Weissenbacher et al. 2002) have led to the reduction of shoaling behaviour in the cave form of G. barreimiae. However, the cave form did not actively avoid conspecifics and still showed a weak preference for the stimulus shoal. There may be some as yet unknown advantage for this behaviour in the cave form, balancing the selection arising from costs of food competition. Future studies will need to investigate the social behaviour of both forms under natural conditions. In summary, we provide an example of a reduction of shoaling behaviour during the evolution of a cave form by a previously surface-dwelling species of fish.