Animal Cognition

, Volume 14, Issue 4, pp 607–612 | Cite as

Spatial cognition in zebrafish: the role of strain and rearing environment

  • Rowena Spence
  • Anne E. Magurran
  • Carl Smith
Short Communication


Two strains of zebrafish, WIK and a second-generation wild strain were reared in either a structurally simple or complex environment and compared in their ability to locate a food reward in a five-chambered maze. There was a significant interaction within subjects between rearing environment and trial, indicating that the consistency of learning varied depending on rearing environment, with those reared in a structurally simple environment showing a slower rate of learning. Fish of both strains reared in a structurally complex environment were smaller than those reared in a simple environment. Our study demonstrates, for the first time in zebrafish, that performance in a learning task as an adult is sensitive to rearing conditions during development.


Cognition Danio rerio Development Environmental enrichment Welfare Learning 


Recent research has provided compelling evidence that fish have significant cognitive ability (Brown et al. 2006). Environmental complexity has long been known to affect cognitive ability in vertebrates (Hebb 1949) through increased neurone density in the hippocampus or analogous brain structures (Kempermann et al. 1997; van Praag et al. 2000; Rosenzweig and Bennett 1996). In hatchery-reared fish, environmental complexity has been shown to enhance post-release survival as a result of improved foraging efficiency and anti-predator behaviour (Berejikian et al. 1999; Brown et al. 2003; Maynard et al. 1995). However, there are few studies that address possible effects of environmental complexity during development on the performance of laboratory fish during experimental procedures.

Here, we used zebrafish, Danio rerio, to test the effects of the structural complexity of rearing environment on adult learning. The zebrafish is one of the most important vertebrate model organisms in genetics, developmental biology, neurophysiology and biomedicine and the most widely used species of laboratory fish (Spence et al. 2008). It is also increasingly utilised in behavioural studies, such as investigations into the effects of drug exposure on learning and memory (e.g. Levin and Chen 2004; Swain et al. 2004). In nature, zebrafish are typically found in structurally complex vegetated, ponds with low water transparency (Spence et al. 2006a), whereas in captivity they are usually kept in bare aquaria under high illumination (Lawrence 2007). If zebrafish are to be used as a model for cognitive behavioural research, the potential for rearing environment to influence cognition needs to be better understood.

Research on zebrafish almost exclusively utilises domesticated strains, many generations removed from wild populations (Spence et al. 2008). However, there is evidence that wild and domesticated strains may differ in some aspects of their behaviour, such as shoal cohesion and boldness (Wright et al. 2003, 2006), while not in others, such as mating behaviour (Spence et al. 2006b, 2007). These different responses may be context dependent; selection for predator avoidance behaviour is likely to be relaxed in captivity, whereas sexual selection may continue to be strong. We used two strains of zebrafish in our study, one a recognised domesticated strain and the other second-generation offspring of wild-caught fish. Rearing environment was manipulated by raising fish of each strain from hatching in either a structurally simple or complex environment. We predicted that the strain with the longer history of domestication would show a lower rate of learning as a result of more relaxed selection for learning compared to a recently derived wild strain. In addition, we predicted that zebrafish raised in a structurally complex environment would show more rapid learning through enhanced neuronal development resulting from navigating through a complex environment during development.



We used two strains of zebrafish in the experiment; WIK, obtained as eggs from the Tübingen zebrafish stock centre and second-generation offspring of wild-caught fish from Bangladesh (hereafter referred to as ‘wild’). The parental generation of the wild fish were collected from sample site 17 in Spence et al. (2006a) in January 2005 and comprised 200 fish. Zebrafish in this population are at risk from both fish and bird predators (Spence et al. 2006a). First and second-generation wild fish were generated from multiple, group spawnings and experimental fish were derived from the mating of a large and outbred population (Spence and Smith 2006). Experimental fish were haphazardly selected as embryos from first generation group matings.

Rearing conditions

We reared fish of each strain at one of two levels of structural complexity. The ‘simple’ environment comprised a bare aquarium, and the ‘complex’ environment an aquarium containing 50 haphazardly placed 50 mm lengths of artificial Elodea canadensis. Therefore, there were four treatment groups: WIK complex, WIK simple, wild complex and wild simple, with 10 independent replicates of each. We allocated 5-day-old embryos randomly to treatments and other conditions (density, temperature, light cycle, water quality) were identical among treatment groups. Test fish were raised in groups of 12 in 20-l glass aquaria (40 × 30 × 25 cm) in an environmentally controlled room on a recirculating system at 25°C with a 14:10 h light: dark cycle and fed twice each day with commercial flake food, placed in a floating feeding ring attached to the wall of the aquarium. When the fish were 6 months old they were isolated in experimental groups comprising three randomly selected individuals drawn from the original group of 12, with the result that experimental groups comprised familiar individuals. Learning rate was tested in a five-chambered maze. Prior to testing, and between periods of testing, fish were housed in bare 20-l glass aquaria. The integrity of experimental groups was maintained throughout the test period to negate any possible effects of familiarity on performance.

Measuring learning rate

Learning was quantified using maze apparatus based on a design by Brown and Braithwaite (2004) and comprising a glass aquarium measuring 50 (width) × 50 (length) × 30 (depth) cm. A central compartment (30 × 30 cm) with opaque walls was connected to four outer compartments by separate 5 cm wide openings (Fig. 1). The aquarium was connected to a recirculating system at 25°C filled to a depth of 18 cm. Water flowed into the central compartment through a pipe and out through each of the outer compartments at the same rate, thus there was a continuous unidirectional flow from the centre of the maze outwards which prevented fish using olfactory cues to locate the reward compartment. The reward compartment held a patch of flake food contained within a floating plastic feeding ring attached to the side of the aquarium. We placed a red coloured marker at the entrance to this compartment as a landmark; see Cameron (2002) for details of zebrafish colour vision. Fish were tested in groups of three; pilot studies showed that isolated fish, and even pairs, sometimes froze on the bottom rather than swimming freely and exploring the maze. Thus, a total of 40 independent groups, each of 3 fish, were tested in trials. The experimental population was male biased, though most experimental groups contained at least one female. Fish were sexed before trials began and the effect of sex on performance was tested.
Fig. 1

Diagram of maze apparatus showing release cylinder (a) in central compartment and feeding ring (b) and marker (c) in reward compartment

Testing procedure

The test fish were transferred to a clear plastic release cylinder in the central compartment of the maze and allowed at least 2 min to settle. The cylinder was then raised remotely enabling the fish to explore the maze. We recorded the time taken for the fish to locate the foraging patch and begin feeding. The fish were tested once each day for 7 days. They were not fed prior to testing and so were motivated to locate the feeding ring. Fish were allowed to feed for 2 min before being returned to their holding aquarium. Each trial continued until at least one fish had commenced feeding, though in all cases at least two fish located the food. If all three fish had not commenced feeding after 10 min, the trial was terminated. After completing seven trials (trials 1–7), fish were retained as a group of three for a further 10 days and fed twice each day. After this period, they were re-tested on a further three consecutive days (trials 8–10) following the original protocol. After completion of 10 trials, fish were measured for body length (BL; tip of the snout to the base of the tail fin) and housed in 60-l aquaria in groups of 20–30 fish and were not used for further experimental work.

Data analysis

Data were loge transformed to meet assumptions of normality and homoscedasticity and the effects of strain and rearing condition on time to feed (feeding latency) were tested using a repeated-measures ANCOVA, with trial as the repeated measure and BL as a covariate. Each experimental group of three fish was treated as an independent replicate; all three fish did not always commence feeding within each 10 min trial so we used the mean feeding latency of the first two fish to feed as the dependent variable, since at least two fish fed in every trial. We used paired t-tests to compare feeding latency before and after the 10-day break in trials. A one-way ANOVA was used to compare fish BL in relation to the sex ratio of experimental groups, and a two-way ANOVA to test for a difference in BL between strains and rearing conditions. A significance level of α = 0.05 was used.


There was a significant difference in fish body length between rearing conditions (two-way ANOVA, F 1,36 = 5.94, P = 0.020) and strain (F 1,36 = 34.45, P < 0.001). There was no significant interaction between treatments (F 1,36 = 0.22, P = 0.640). WIK were larger than wild fish, while fish reared in a simple environment were larger than those raised in a complex environment. Mean (±SD) standard lengths were: WIK complex, 25.4 (±1.79) mm; WIK simple, 26.1 (±2.06) mm; wild complex 22.9 (± 1.54) mm; wild simple 24.0 (±1.73) mm. There was no correlation between BL and mean feeding latency over the first seven trials for wild fish (Pearson correlation, r 18 = −0.31, P = 0.187), for WIK fish there was a negative correlation (Pearson correlation, r 18 = −0.61, P = 0.004). There was no difference in BL between sexes (unpaired t-test, t 118 = 0.05, P = 0.963), neither was there any effect of sex ratio of test group (i.e. whether the group contained one, two or three males) on latency to feed (one-way ANOVA F 1,36 = 1.73, P = 0.191).

Between treatments, fish BL was a significant covariate (Table 1). Although wild fish took longer than WIK to commence feeding at the start of trials (Fig. 2), after adjusting for BL, there was no significant effect of strain (Table 1). There was also no overall significant effect of rearing environment, or interaction between rearing environment and strain (Table 1).
Table 1

Repeated measures ANCOVA results for the performance of zebrafish in a five-chambered maze task


Trials 1–7

Trials 8–10





Between subjects

 Fish length










 Rearing environment





 Strain × rearing environment





Within subjects






 Trial × fish length





 Trial × strain





 Trial × rearing environment





 Trial × strain × rearing environment





Significant results are in bold type

Fig. 2

Mean feeding latency (s) of first two fish over 10 trials in each of four treatments: a WIK complex; b WIK simple; c wild complex; d wild simple. Arrows indicate the re-commencement of trials after a ten-day break

Within subjects there was a significant decrease in latency to feed over the initial 7 trials (Table 1, Fig. 2). There was no interaction between zebrafish strain and trial (Table 1), showing that the rate of improvement between strains among trials was comparable. However, there was a significant interaction between BL and trial, and rearing environment and trial (Table 1). These results indicate that both fish size and rearing environment influenced latency to feed. Larger fish showed a shorter latency to feed than smaller individuals, and those reared in a complex environment showed a different pattern of learning to those reared in bare tanks. For the former result, partitioning the interaction by means of orthogonal polynomials showed a significant linear effect (F 1,35 = 9.20, P = 0.005), but no higher order effects (P > 0.10). For the rearing environment by trial interaction, there were significant higher order effects (4th order term: F 1,35 = 6.43, P = 0.016; 5th order term: F 1,35 = 6.91, P = 0.013). There was a significant three-way interaction between strain, rearing environment and trial (Table 1), indicating a different response between strains in the pattern of learning in relation to rearing environment. In this case, partitioning the interaction showed a significant cubic (3rd order term) effect (F 1,35 = 6.03, P = 0.019).

In the first trial following the 10-day break in testing, fish took significantly longer to commence feeding than in trial 7 (paired t-test, t 39 = 3.70, P = 0.001), although they were faster than in trial 1 (paired t-test, t 39 = 3.290, P = 0.002). There was no significant improvement in performance over the 3 trials after the ten-day break (Table 1, Fig. 2), and no interactions between trial and zebrafish strain or rearing environment (Table 1). Between treatments there was no significant strain effect (Table 1) and no effect of rearing environment or interaction between treatments (Table 1).

All the foregoing results are for the first two fish to commence feeding. When these analyses were repeated using data for just the first fish to feed the outcomes were almost identical, with the exception of the three-way interaction between strain, rearing environment and trial over the first seven trials, which was non-significant (F 6,210 = 0.47, P = 0.832).


The goal of this study was to examine the effect of environmental complexity during development on learning in two strains of zebrafish. All four treatment groups demonstrated learning over the initial seven trials, with the most rapid change in feeding latency taking place in the first four trials. Re-testing after a 10-day break showed that learning had started to decay, although performance was still better than at the commencement of testing. There was no improvement in performance with a two further trials. There were wide differences within trials among replicates, though the use of groups of fish helped to buffer these differences; other studies using individual zebrafish have suffered from high variability and have required up to 60 trials per fish (Bilotta et al. 2005; Colwill et al. 2005).

We detected a significant interaction between rearing environment and trial. Fish reared in a structurally complex environment showed a significantly different pattern of learning to those reared in bare tanks, with the latter reaching a minimum more rapidly. In trials 8–10, there was no interaction between rearing environment and trial (Table 1). These results indicate that the rate of learning, at least during the initial phase, was more rapid for fish reared in a structurally complex environment compared to a simple one, irrespective of strain.

The environment in which an individual develops can influence memory and ability to learn. Experimental studies with birds, mammals and fish have demonstrated that learning and orientation behaviour can be directly linked to the type of environment from which they are taken (Carlier and Lefebvre 1997; Maynard et al. 1995; Rosenzweig and Bennett 1996; Wiltschko and Wiltschko 1989). The possible mechanisms by which environment influences memory and cognition may be innate. Alternatively, there may be a plastic response during development in response to specific features of the environment. An additional scenario is that both processes may operate (Kieffer and Colgan 1992). In the three-spined stickleback (Gasterosteus aculeatus), a small territorial fish, individuals from a river population were shown to have a different learning ability, in terms of the cues used for spatial orientation, in comparison with those from an adjacent pond (Girvan and Braithwaite 1998). When cross-reared, fish from the two environments were shown to learn spatial cues relevant to the environment in which they were raised, with little evidence of an innate population effect (Girvan and Braithwaite 2000), indicating that plastic responses were the principal mechanism for population differences in learning. In contrast, guppies (Poecilia reticulata) from populations naturally exposed to different predation regimes showed disparities in spatial learning that appeared to be innate and which correlated with the size of telencephalon (Burns and Rodd 2008). The telencephalon is the region of the brain that comprises the lateral pallium, which is the seat of spatial memory in fish and analogous to the hippocampus in higher vertebrates (Nieuwenhuys 1963; Rodríguez et al. 2002). Notably, we detected no significant difference between zebrafish strains in their rate of leaning. Despite many generations in captivity the WIK fish completed the learning task as quickly as the wild fish. We did not attempt to compare brain structure between the strains we used. However, this result suggests that learning rate may be conserved within this species.

Unexpectedly, fish of both strains were significantly smaller when reared in a structurally complex environment. The effect of increasing environmental complexity may have been to increase overall density and thereby magnify competition for food. Greater structural complexity is often presumed to offset the potentially negative effects of dominance and aggression (e.g. Basquill and Grant 1998), though in zebrafish this assumption may need to be reappraised. Dominance typically correlates with size in fish (Casalini et al. 2009; Wootton 1998), including zebrafish (Spence et al. 2008), and increased environmental complexity might therefore be predicted to increase variance in size among tank mates, possibly leading to elevated aggression.

Our findings have implications for zebrafish welfare and the design of protocols using zebrafish in behavioural research, which are increasing (Spence et al. 2008). Clearly, certain aspects of rearing conditions are potentially confounding factors in behavioural studies that test learning and memory, such as those assessing the effects of drugs of abuse on learning and memory. However, while it is important to control for extraneous effects (Ninkovic and Bally-Cuif 2006), there is also a danger that too great an emphasis on standardisation can result in loss of external validity (Würbel 2000). To date, little effort has been made to standardise rearing and husbandry conditions for laboratory zebrafish, though the problem has at least been recognised (Lawrence 2007).

In conclusion, our study shows that learning in zebrafish can be altered under different rearing conditions, with the structural complexity of the rearing environment during development affecting performance in a learning task as an adult. There was also a significant negative effect of increased environmental complexity on growth rate, which may arise through amplified competition for food in a more structurally complex environment.



We are grateful to Culum Brown and Martin Reichard for their comments on an earlier version of the manuscript and to Robin Wilson for statistical advice.


  1. Basquill SP, Grant JWA (1998) An increase in habitat complexity reduces aggression and monopolization of food by zebra fish (Danio rerio). Can J Zool 76:770–772Google Scholar
  2. Berejikian BA, Smith RJF, Tezak EP, Schroder SL, Knudsen CM (1999) Chemical alarm signals and complex hatchery rearing habitats affect anti-predator behaviour and survival of Chinook salmon (Oncorhynchus tshawytscha) juveniles. Can J Fish Aquat Sci 56:830–838CrossRefGoogle Scholar
  3. Bilotta J, Risner ML, Davis EC, Haggbloom SJ (2005) Assessing appetitive choice discrimination learning in zebrafish. Zebrafish 2:251–268CrossRefGoogle Scholar
  4. Brown C, Braithwaite VA (2004) Effects of predation pressure on the cognitive ability of the poeciliid Brachyraphis episcope. Behav Ecol 16:482–487CrossRefGoogle Scholar
  5. Brown C, Davidson T, Laland K (2003) Environmental enrichment and prior experience of live prey improve foraging behaviour in hatchery-reared Atlantic salmon. J Fish Biol 63:187–196CrossRefGoogle Scholar
  6. Brown C, Laland K, Krause J (2006) Fish cognition and behaviour. In: Brown C, Laland K, Krause J (eds) Fish cognition and behaviour. Blackwell, Oxford, pp 1–8Google Scholar
  7. Burns JG, Rodd H (2008) Hastiness, brain size and predation regime affect the performance of wild guppies in a spatial memory task. Anim Behav 76:911–922CrossRefGoogle Scholar
  8. Cameron DA (2002) Mapping absorbance spectra, cone fractions, and neuronal mechanisms to photopic spectral sensitivity in the zebrafish. Vis Neurosci 19:365–372PubMedCrossRefGoogle Scholar
  9. Carlier P, Lefebvre L (1997) Ecological differences in social learning between adjacent, mixing, populations of Zanaida doves. Ethology 103:772–784CrossRefGoogle Scholar
  10. Casalini M, Agbali M, Reichard M, Konecna M, Bryjova A, Smith C (2009) Male dominance, female mate choice and intersexual conflict in the rose bitterling (Rhodeus ocellatus). Evolution 63:366–376PubMedCrossRefGoogle Scholar
  11. Colwill RM, Raymond MP, Ferreira L, Escudero H (2005) Visual discrimination learning in zebrafish (Danio rerio). Behav Proc 70:19–31CrossRefGoogle Scholar
  12. Girvan JR, Braithwaite VA (1998) Population differences in spatial learning in three-spined sticklebacks. Proc R Soc Lond B 265:913–918CrossRefGoogle Scholar
  13. Girvan JR, Braithwaite VA (2000) Orientation behaviour in sticklebacks: modified by experience or population specific? Behaviour 137:833–843CrossRefGoogle Scholar
  14. Hebb DO (1949) The organisation of behaviour. Wiley, New YorkGoogle Scholar
  15. Kempermann G, Kuhn HG, Gage HF (1997) More hippocampal neurons in mice living in an enriched environment. Nature 386:493–495PubMedCrossRefGoogle Scholar
  16. Kieffer JD, Colgan PW (1992) The role of learning in fish behaviour. Rev Fish Biol Fish 2:125–143CrossRefGoogle Scholar
  17. Lawrence C (2007) The husbandry of zebrafish (Danio rerio): a review. Aquaculture 269:1–20CrossRefGoogle Scholar
  18. Levin ED, Chen E (2004) Nicotinic involvement in memory function in zebrafish. Neurotoxicol Teratol 26:731–735PubMedCrossRefGoogle Scholar
  19. Maynard D, Flagg T, Mahnken C (1995) A review of semi-culture strategies for enhancing the post-release survival of anadromous salmonids. Am Fish Soc Symp 15:307–314Google Scholar
  20. Nieuwenhuys R (1963) The comparative anatomy of the actinopterygian forebrain. J Hirnforsch 13:171–192PubMedGoogle Scholar
  21. Ninkovic J, Bally-Cuif L (2006) The zebrafish as a model system for assessing the reinforcing properties of drugs of abuse. Methods 39:262–274PubMedCrossRefGoogle Scholar
  22. Rodríguez R, López JC, Vargas JP, Gómez Y, Broglio C, Salas C (2002) Conservation of spatial memory function in the pallial forebrain of reptiles and ray-finned fishes. J Neurosci 22:2894–2903PubMedGoogle Scholar
  23. Rosenzweig MR, Bennet EL (1996) Psychobiology of plasticity: effects of training and experience on brain and behaviour. Behav Brain Res 78:57–65PubMedCrossRefGoogle Scholar
  24. Spence R, Smith C (2006) Female mate preferences in the zebrafish (Danio rerio). Behav Ecol 17:779–783CrossRefGoogle Scholar
  25. Spence R, Fatema MK, Reichard M, Huq KA, Wahab MA, Ahmed ZF, Smith C (2006a) The distribution and habitat preferences of the zebrafish in Bangladesh. J Fish Biol 69:1435–1448CrossRefGoogle Scholar
  26. Spence R, Jordan WC, Smith C (2006b) Genetic analysis of male reproductive success in relation to density in the zebrafish, Danio rerio. Front Zool 3:5PubMedCrossRefGoogle Scholar
  27. Spence R, Ashton RL, Smith C (2007) Adaptive oviposition decisions are mediated by spawning site quality in the zebrafish, Danio rerio. Behaviour 144:953–966CrossRefGoogle Scholar
  28. Spence R, Gerlach G, Lawrence C, Smith C (2008) The behaviour and ecology of the zebrafish, Danio rerio. Biol Rev Camb Philos Soc 83:13–34PubMedGoogle Scholar
  29. Swain HA, Sigstand C, Scalzo FM (2004) Effects of dizocilpine (MK-801) on circling behaviour, swimming activity, and place preference in zebrafish (Danio rerio). Neurotoxicol Teratol 26:725–729PubMedCrossRefGoogle Scholar
  30. van Praag H, Kempermann G, Gage FH (2000) Neural consequences of environmental enrichment. Nat Rev Neurosci 1:191–198PubMedCrossRefGoogle Scholar
  31. Wiltschko R, Wiltschko W (1989) Pigeon homing: olfactory orientation—a paradox. Behav Ecol Sociobiol 24:163–173CrossRefGoogle Scholar
  32. Wootton RJ (1998) Ecology of teleost fishes. Kluwer, DordrechtGoogle Scholar
  33. Wright D, Rimmer LB, Pritchard VL, Krause J, Butlin RK (2003) Inter and intra-population variation in shoaling and boldness in the zebrafish (Danio rerio). Naturwissenschaften 90:374–377PubMedCrossRefGoogle Scholar
  34. Wright D, Nakamichi R, Krause J, Butlin RK (2006) QTL analysis of behavioural and morphological differentiation between wild and laboratory zebrafish (Danio rerio). Behav Genet 36:271–284PubMedCrossRefGoogle Scholar
  35. Würbel H (2000) Behaviour and the standardization fallacy. Nat Genet 26:277–281CrossRefGoogle Scholar

Copyright information

© Springer-Verlag 2011

Authors and Affiliations

  1. 1.School of BiologyUniversity of St. AndrewsFifeUK

Personalised recommendations