supervised by Roger SchOrch & Dik Heg internal supervisor Theunis Piersma
I I I I
D96
Fardo Witsenburg 16 — 07 — 2008
Puttina Dersonalutv back in nature:
Natural variation in behavioural types in the
cooperative breeding cichlid Neolamprologus puicher
a master research report
I I I
I I I
I I I I I I
I
I
I
I
CONTENTS
Abstract 3
Introduction 4
Behavioural syndromes 4
Cooperative behaviour 6
Hypotheses 7
Methods 9
Study species and site 9
Assessing behavioural type 11
Data management & statistical methods 12
Results 14
Consistency over time 14
Correlations among behavioural traits 14
Principal component analysis 14
Component correlates 17
Discussion 21
Main findings 21
Consistency and inter-correlation 22
The other components 22
Measurement considerations 23
Acknowledgements 25
References 26
Appendix A 30
2
-u
ABSTRACT
Consistent inter-correlated variation of behaviours, known as a 'behavioural syndrome' has been identified in populations of a broad range of animal species. Still little is known about and the ecological relevancy of behavioural syndromes. In a system of cooperative breeding behavioural type might critically influence an indivIdual's decision to help or to disperse.
Neolamprologus puicher Is a cooperatively breeding cichlld and faces many important life- history decisions about dispersal and helping effort that have to be traded off against each other. We assessed the behavioural type of size matched subordinates in their natural environment by quantifying their aggression against a conspecific, anti-predator behaviour, roaming behaviour and helping effort.
A principal component analysis grouped all behaviours but one In the first component, demonstrating strong co-variation among these behaviours (a behavioural syndrome).
Surprisingly, roaming was negatively correlated with the other behaviours like aggression.
Digging, as a measure of help, did not load Into this component and its rank orders were not maintained over time. Other behaviours varied in their consistency across time, but all showed a positive trend. The sexes did not differ In their behavioural type and the behavioural types did not have different feeding rates. In larger breeding groups, subordinates were more aggressive and helpful and less exploratory. This demonstrates the trade-off between helping the current group and looking for independent breeding opportunities. Location In the colony, predator density nor competitor density influenced the distribution of behavioural types.
This study demonstrates that natural variation in co-varying behaviour indeed occurs.
Helping behaviour varies with the behavioural types and should therefore have consequences for the entire life-history.
3
INTRODUCTION
Behavioural syndromes
In 1976, Felicity Huntingford found that seemingly identical sticklebacks differed in their behaviours consistently throughout their breeding cycle (Huntingford 1976). Personality, once thought to be the exclusive domain of humans and some mammalian and avian relatives, was suddenly a property to be found in rather distant taxa. Huntingford herself did not use the term 'personality' in her paper from 1976 but only mentioned consistent and co-varying behaviour levels. The field of animal personality has taken flight since then, identifying correlated traits in taxa ever more distant from humans (Smith & Blumstein 2008). For example in the fishing spider Dolomides triton the most voracious young grow up to be the most voracious adults of their generation. Furthermore, these same voracious individuals have a higher likelihood of eating potential partners before mating (Chadwick Johnson & Sih 2005).
The idea of correlated behavioural traits has formalised in the framework of 'behavioural syndromes' which we define as "a suit of correlated behaviours reflecting between-Individual consistency In behaviours across multiple situations (Sih eta!. 2004). A 'behavioural type' is the collection of values of behavioural traits for one individual. The behavioural type is therefore the property of an individual, whereas a behavioural syndrome is the property of the entire population (Bell 2007). This terminology may still differ between papers, stating definitions therefore remains important; e.g. Smith and Blumstein (2008) seem to use the term 'behavioural syndrome' for our definition of 'behavioural type'.
One of the consequences of a behavioural syndrome is (maladaptive) spill-over effects; for example the pre-copulatory sexual cannibalism in the fishing spider (Chadwick Johnson & Sih 2005). Consistency in behavioural traits was therefore considered to be some
sort of constraint; these correlated responses could only arise through either plelotropy or linkage disequilibrium (Price & Langen 1992). Why else would an animal deliberately fixate its behaviours, which, up to then, were always considered to be infinitely plastic? If physiological or ontological processes cause conflicts between behaviours, preventing them from reaching
Fig. 1. Definitions of behavioural type and behavioural syndrome. Each data point represents a different individual in the population. From Bell (2007)
4 bthao X
their optimal level, selection would be expected to decouple these processes from each other (Slh eta!. 2004). Plasticity was thought to be so costly that Individuals were better off following the same strategy continuously (DalI eta!. 2004).
New discoveries lead to the view that behavioural syndromes areIndeed adaptive; in ponds with predators sticklebacks have evolved a syndrome In which boldness and aggression are correlated. In predation free ponds the stickleback populations lack this correlation (Dingemanse eta!. 2007). When a predator-naive population (without thesyndrome) Is confronted with a predator, the population is driven towards the behavioural correlation; but the process is not only caused by correlational selection, some individuals will actively adapt their behaviour (shy individuals reduce their aggression levels) to fit the syndrome (Bell & Sih 2007). Adopting a 'fitting' behavioural type, whether actively or by selection,therefore seems to be an adaptive response.
But why are there different 'fitting' behavioural types within oneecological context?
I.e. why Is there a continuum of personalities and not just one type adopted by everyone? The general theories on the preservation of variation could be applicable (temporal or spatial variation In conditions, frequency dependent selection, mutation and weak selection; Sih eta!.
2004), though these models do not explain the within-individual consistencyof behavioural variation (Stamps 2007). When being predictable in one's behaviour is important (e.g. In hawk-dove games with eavesdropping), the evolution of consistent traits will be favoured (Dali eta!. 2004) but it fails to explain the correlation between traits.
In a model by Wolf et a!. (2007) indivIdual differences In fitness expectation cause the evolution of consistent and correlated variation in behaviour and thusthe development of a behavioural syndrome. Individuals could choose to reproduce immediately or investIn exploring, thereby increasing there reproductive output for the second year.Between the two reproductive events, all were confronted with several risk-taking and hawk-dove games that provided a payoff when won. The reproductive strategy and both game tactics were
genetically determined and all combinations were possible, yet the model showed that evolution favoured the development of correlated traits (Wolf eta!. 2007). Two distinct behavioural types evolved; one risky type that reproduced immediately and took high risks in
both games. The other type was averting all risk but invested in thesecond reproduction. A life history difference (the reproductive strategy In this case) determines the types of other behaviour shown by an Individual.
This same idea has been proposed by Stamps (2007). She hypothesizesthat a trade- off between growth and survival will cause behaviour patterns that reinforce either growth or survival. Behaviours will thus be correlated. Strengths of this theory are that all behaviours can be evaluated using this trade-off and the direction of correlation can be predicted depending on whether the behaviour increases growth (e.g. exploring new food items) or survival (e.g. hiding from predators).
5
That growth rate determines the behavioural type of an individual might be a new idea, but the reverse, behavioural type influencing fitness, has been considered much longer.
Reproductive output and survival are both expected to be Influenced by the behavioural type of an individual. A meta-analysis tried to identify the general effects on these traits but found only boldness to be of influence on both reproductive success and survival whereas exploration and aggression gave more ambiguous results (Smith & Blumsteln 2008).
Other life-history traits are also under influence of a behavioural syndrome.
Attractiveness to mates in guppies (Godln & Dugatkln 1996), food differentiation in brook chaff (Wilson & McLaughlin 2007), dispersal distance in great tits (Dingemanse eta!. 2003)
and mating strategies In lizards (Stapley & Keogh 2005) are all (at least partially) correlated with the personalIty of the carrier. A full picture of the effects of personality on the total life- history is as of yet not available but would give us Insight Into the mechanism of selection on behavioural type (Reale eta!. 2007).
Cooperative behaviour
Why some species are cooperative breeders and others are not remains unresolved. In fact, this question can be separated into two (Bergmuiler eta!. 2005b): Why are Individuals not dispersing and why do they decide to help? Ecological constraints were thought to Inhibit dispersal, due to costs associated with leaving the parental nest or saturation of the habitat.
Longevity and a level of high parental care were found to promote delayed dispersal aswell (Hatchwell & Komdeur 2000). Theoretical analysis revealed that If delaying dispersal is associated with direct benefits (e.g. by Inheriting the territory), having these direct benefits Is a better predictor of a cooperative system than the presence ofecological constraints or certain life-history traits (Pen & Weissing 2000). In all cases it is assumed that delayed dispersal will somehow lead to helping behaviour (e.g. through kin selection or as payment under threat of eviction; "pay-to-stay").
T
Fig. 1. SchematicI' '\
representation of the range of social decisions available during an individuals lifetime. Solid lines connect the decision trajectory with the resulting breeding strategy. From Cahan et a!. (2002)6
The theory of kin selection as an explanation of helping behaviour (Hamilton 1964) has lost a lot of its explanatory power outside the haplo-dipiolds in the last decades; several genetical analyses revealed that within social groups relatedness is not higher compared to uncooperative groups (Clutton-Brock 2002,Komdeur 2006). When helpers are unrelated to the recipients they serve, the indirect benefits of helping will be non-existent, but the direct benefits should play a much larger role. Mutualistic reciprocity, mutualism as a by-product of fitness maximization (e.g. by gaining breeding experience) or the benefits of group
augmentation have all been identified as possible causes for helping others (Clutton-Brock 2002).
Apart from evolutionary perspective, the question to help or not can also be scaled down to the level of the individual. This dilemma can be separated into several smaller decisions, each made In the light of fitness maximization (Cahan eta!. 2002). This breakdown of decisions does not only clarify the decision-making process but also shows the
consequences of each choice for the life-history of the individual (see fig. 2). If individuals differ in their decision making due to some underlying variation, this will have severe
consequences for their breeding position and therefore the genetic make-up of the population.
Hypotheses
Helping is a type of behaviour (or set of behaviours), no different from other behaviours and individuals will therefore vary in their level of helping. Indeed, not only has variation in helping propensity been found, but it was also demonstrated to be consistent in several populations (Komdeur & Edelaar 2001,Wemer eta!. 2003,Amold eta!. 2005). This consistent variation in helping effort might very well be determined by some underlying physiological variation, in analogy to the variation in behavioural type (Stamps 2007). There is in fact no reason not to assume that all these behaviours are determIned by the same underlying trait. Helping
behaviour would therefore co-vary with other known consistent behaviours like aggression and exploration. Here we test the hypothesis that helping behaviour is part of this behavioural syndrome.
Activity, aggression, boldness and exploration often co-vary and are sometimes referred to as an activity or risk-taking syndrome. Risk-taking behaviour would be associated with leaving the safety of the territory and exploring around. Individuals showing a lot of these behaviours would not be suitable helpers since they would not be around as often to help.
More likely will these individuals disperse instead; helpers that invest little In their help have been found to have a high chance of dispersing (Bergmuller eta!. 2005b). The less active and risk-taking individuals are therefore expected to be the more providing helpers.
The relation between behavioural type and helping propensity has been looked into by Schürch and Heg (2008). In the cooperatively breeding cichlid Neolampro!ogus pu!cher they found evidence for a behavioural syndrome consisting of boldness, aggression and exploration, all positively correlated. In females these behaviours correlated positively with sand carrying
7
(a behaviour considered as helpful) but this relation was not found in males. In the Zebrafish Danlo rerio it was shown that lab-strains and wild-strains differ in their behavioural
correlations, demonstrating in fact a domestication-syndrome (Moretz eta!. 2007). Findings from the lab can therefore not be immediately projected on populations In the wild. Moreover, the authors themselves questioned their measures of traits, stating that the strength of a relationship depended on how the trait was defined. The ecological relevancy of tests remains a point of concern In behavioural sciences (Reale eta!. 2007). Almost all research on
behavioural syndromes comes from lab studies. Field studies on the ecology of behavioural syndromes are rare but vital. How does the behavioural type of an individual help that individual in surviving in its environment? Answering these kind of questions will critically improve our understanding of the evolution of personalities.
Our study therefore focused on a natural population of Neo!ampro!ogus pu!cher.
Aggression, anti-predator behaviour, exploration and helping behaviour of helper fish were scored in their natural, but experimentally manipulated, environment. We used this data to look for a behavioural syndrome and how helping relates to It. Furthermore, we hoped to answer questions on how this behavioural variation was distributed through the population. Is there a difference between the sexes? How does the behavioural type relate to growth?
Variation In growth rate had already been mentioned as a good underlying mechanism determining the risk-taking behavioural type (Stamps 2007). Is there a spatial pattern to be found? A spatial pattern could be present based on differences In exploration, finding better quality breeding groups to join when an individual explores his surroundings more. And how are behavioural types distributed among breeding group sizes? A larger group, with more helpers, would require less help per helper. The answer to these questions would expand our knowledge of behavioural syndromes into their natural surroundings.
8
I I I
I
MATERIAL AND METHODSStudy species and site
I
The study was performed on a population of Neolamprologus puicher (Pisces, Cichlidae), a cichlid endemic to Lake Tanganyika. It is 1 of the 19-2 1 cooperative breeding cichlid species in'
Lake Tanganyika (Heg & Bachar 2006). A dominant male breeds with one to several females within a defended territory, consisting of a few stones providing shelters. A colony consists of many of these clustered territories that share common boundaries (Balshine-Eam eta!. 1998).I
The breeding pair receives help from sexually mature helpers who themselves refrain from breeding but instead help fanning the eggs, protecting fry, cleaning and digging out shelters and maintaining the territory. Groups can include up to 20 of these subordinates (Balshine et a!. 2001,Heg eta!. 2005) and because helpers tend to stay when there is a switch in breeders, average relatedness decreases with age of the helper (Dierkes eta!. 2005). Helpers provideI
less help when breeding opportunities become less restricted, until they disperse (Bergmuller et a!. 2005b). Within a colony both sexes disperse equally, but only large males disperse further away to other colonies (Stiver et a!. 2007).
We studied two colonies of N. pulcher 400 m offshore at Kasakalawe point, near the Zambian town of Mpulungu at the southern tip of Lake Tanganyika. N. puicher breed here at a depth of> 9 m on a rocky or pebbled substrate interspersed with sandy stretches. Throughout the study period (30-09-2007 to 07-12-2007) preparations and observations were done scuba diving.
Of all 124 breedIng groups within the study site the family compositions were visually I assessed and their exact location determined using a 2x2 m grid covering the whole area of
interest (see fig. 3). 70 groups were marked using numbered stones flagged with a floating 5 I ml eppendorf cup. Of the 70 thus marked groups 30 indivIdual helper fish were caught using a I transparent acrylic glass tube and hand net. These helpers were measured for standard
length, visually sexed by their gonadal pore and marked with clips In their dorsal and anal fin.
All handling of the fish was done at the study site. The caught fish were released immediately after marking and were not disturbed for the next 24 hours. Within 3 to 4 weeks the fins would recover from the clippings making the markings undetectable. It was therefore I necessary for most helpers to be remarked once during the testing period, using the above
described protocol. When all the observations had finished, all focal individuals were
I
recaptured using gillnets and measured a third time.
To estimate local predator and space competitor density at our study site we used two different approaches. On three dives we observed every 2 x 2 m square of our grid for 90
I seconds and counted the number of indivIduals for 21 fish species that were present. These species are predators of N. pu!cher or compete for breeding space with them (for a list of species see appendix A). Therefore, we mapped their breeding territories in the same way we
1 did with the N. puicher shelters (see fig. 3).
1 9
I
I
0 0
o
•o •
00 0 0
o
0
0o 0
00
•
0
0
0 0
0
0
0
00 0
0 0
a.
0•
o
•
0
• •. S. • .
OS • 05
0•0 .00
0• . •0
0
•5•
o
•
0OoO
0.
oS •
0
00
.0.. S! 0 0
0
iS • •
5.%0i0
0
0055
0
o
0
•
5Oo
0 0 00. •
0
00
0•O .0
0 00
• •
000 0
Fig. 3.Map of the study-site. Dashedlines are the 2x2mgrid, used for orientaon.
Black dots are territories ofN. puicher, white dots are territories of all other fish.
TwoN. puicher colonies are disbnguishable; a small colony in the bottom left, and a large one in the top half of the map.
10
S I I
I :
I
'
I L
S
-d- .
S
.
I
I I I 'I,
.
S
-
a.
Fig. 4. Schematic representations of density calculations. Thin solid lines are the 2 x 2 m grid, a) Density of space competitors is determined by countin9 all territories (blue dots) within a virtual square around the focal territory (red star). Large scale density: large dashed box, small scale density: small dotted box. b) The predator density for the focal territory (red star) is determined by the overlap of four squares by the 2x2m virtual square (dashed box). As the largest part of this box overlaps square 2, its predator density is weighted heaviest. Square 3 has the smallest surface overlap and will contribute the least to the focars predator density.
From these data we calculated a measure for local space competitor density by counting territories of all species around the focal's home territory in a 1 x 1 m square (small scale local density) and in a 4 x 4 m square (large scale local density). Local predator density for every focal's territory was calculated for a 2 x 2 m square around the home territory. The overlap of this virtual square with the counted 2 x 2 m squares of the grid was used to calculate a weighted average of their densities, based on their amount of overlap (see fig. 4).
Assessing behavioural type
From the 30 caught-and-marked fish 19 individuals were selected for behavioural type analysis, based on equal sex and space distribution. Four individuals disappeared during the study period, all male. The complete test series were therefore performed on 15 helper-fish, with a 5:10 male:female sex-distribution. To test for consistent behaviour over different contexts these helpers were exposed to three different treatments in their natural
environment. A transparent acrylic glass presentation tube, with mesh wire bottom was placed in the middle of the focal's breeding group territory. The tube contained either nothing, a competitor (size range 38-54 mm) or the natural predator Lepidiolamprologus elongates (Pisces, Cichlidae, size range 59-95 mm) both caught outside the study area and kept between trials in underwater cages up to 2 (N. puicher) and 4 days (L. elongatus).
After placement of the presentation tube the position and behaviour of the focal, together with its number of feeding bites, was scored for 10 minutes by direct observation.
The position was noted down as the nearest territory and its position to that territory
I I
I I
I
S
S S S
I
S
2m
I
m
I I I
b.
.1
I
I I I I I I I
11
('nearby', 'on the territory' or 'in a shelter'). On this position aggressive (biting, bumping, chasing, charging, spreading fins, spreading gills, head down) and digging (sand carrying, digging) behaviours were noted down, together with the actor and receiver (conspecific, predator, other heterospecific or the tube) of that behaviour.
The three different treatment tubes (nothing, competitor, predator) were presented consecutively in random order, resulting In a set of 3x10 minutes of observations. If the focal had gone to far from the breeding territory to see the tube, the observer would wait with placing the new tube until It had returned to the territory.
Every focal helper was observed for four times, with Intervals of 5 to 15 days (modus:
8 days), resulting in a total of 4x3x10 mm = 120 mm of observational data per individual. All observations were conducted 1-2 meters from the territory by W. F. Witsenburg. Observations started Immediately after placing the treatment tube, 1-2 mInutes after the observer had spotted the marked individual at the site.
Data management & statistical methods
Four main behaviours were tested for their co-variance: 'aggression against competitors', 'anti-predator behaviour', 'explorative behaviour' and 'helping'. Some of these behaviours
were quantified by several measures. Aggression against a competitor consisted solely of the number of aggressive acts performed by the focal towards the tube when it contained a conspecific. Similarly, anti-predator behaviour was quantified by the number of aggressive acts performed towards the tube when the predator L. elongatus resided in the tube. To cover the whole spectre of explorative behaviour, three different measures were used; the number of times that the focal left its territory, the number of unique territories it visited and the
maximal distance it went away from the territory during the ten-minute observation. Helping behaviour was quantified by counting the number of digging behaviours performed by the focal on its territory when the empty tube was present and by frequency of visits to a shelter on the territory.
Conditions In the field could change rapidly; To get rid of the noise this may have caused In our data, we averaged same type observations of the same individual. To test for consistency over time of behavioural traits, we compared the average of that trait of the first two observations with the average of the last two observations. As the variances of the averaged traits were not normally distributed, and transformations did not help in all cases, we used Spearman's rank correlation. For correlation between traits, trait averages of all four observations were used. Significance of correlations was tested using Spearman's rank correlation.
To Increase the effectiveness of the principal component analysis, 5 out of 7
measurements were transformed. Aggression against the competitor and aggression against the predator were both cube-root transformed. The number of times that the focal left its territory, the number of unique territories it visited and the number of digging behaviours
i2
performed were all square-rooted. After transformation the co-variance of the behavioural measurements were summarized with a principal component analysis using single value decomposition on centred and scaled values. Verification by repeated analysis with random subsamples was not possible due to small sample size; therefore a scree-plot was used to identify the significant components.
The significant principal components were used for life history and ecology analysis.
The number of feeding bites used in this analysis was the average of the four observations with an empty tube. The relative location of a territory within the colony was visually
determined. Growth was left out of the analysis; the size measurements were taken on a too coarse scale to detect accurate effects. All correlations were tested using Spearman's rank correlation. Categorical comparisons were done using Student's t test or Welch t test, depending on the variances.
13
RESULTS
Consistency over time
For all quantified traits the observed variance was not normally distributed. Of all the tested traits only aggression against a competitor, shelter visits and departures from the territory were significantly consistent over time. For most other traits there was a tendency for the average of the first two trials to have a positive relation with the average of the last two trials (see fig. 5 and table 2). Digging behaviour was an exception, the first two observations not having any predictive power over the last two.
Correlations among behavioural traits
Table 1 shows the co-variance of the behavioural measures. Not surprisingly, the three spatial measures (number of unique territories visited, maximum distance away from territory and number of times away from own territory) were strongly correlated. The significant correlation between aggression against a predator and these spatial measures was less Intuitive and already showed that a wild population of N. puicher has suits of correlated behaviours.
Principal component analysis
The scree-plot revealed that out of the possible seven components, three principal
components provided most of the explanatory power, of which two had an eigenvalue larger than 1 (see table 4). All factors (the transformed behavioural measurements) loaded considerably into the first component, with the notorious exception of digging (table 3). The sign Indicates that aggression and visiting the shelter were all In the same way related to the first component and therefore positively related to each other, but negatively related to the three spatial measures (who in their turn were positively related to each other). Therefore the first PC showed to us a suit of several correlated behaviours; it reflects a syndrome in which more roaming individuals are less aggressive and help less. This syndrome-component explained over 50% of our observed variation In behaviours.
The second component accounted for the variation caused by digging behaviour and the strongly correlated shelter visiting. Another 14% of the variation In the data was
accounted for by the third component. This component was most strongly correlated with the two forms of aggression and, to a lesser extent, with the number of unique patches visited.
The relation between aggression and the number of visited patches was positive, in contrast with the first component.
14
— — — — — — — — — — — — — — — — — — — — —
SSS aS SS SviS g•..4_ I•I :'
• I
• IS 9S •SI•
.•
ISS •2eaS S...
S e• IIIIIIIIIIIIIIII os10IS202530051015201.015202.53.0020406080 VSfS9S I a'°'
bavarag1cavIrag.d aggrnvion towards coinp.titaraggression toward. predatoruniqits Iwiloii.s visd.dmaxwin.n dátanc. ras1.d w an S•SI0. I
•: I.
I •
airags Iaverage IlyIng. I # es asisiy tram twermrye• siisttar v*itif
digmgaction.g Fi9. 5. Consistency over time of the 7 behavioural traits. x-axis shows the average value of the first two observations, the y-axis shows the averageof the last two observations. Values are untransformed, a) aggression towards the competitor; b) aggression towards the predator; c) # unique territories visited; d) maximum distance travelled from home temtory; e) # departures from home territory; t) # visits to the shelter; g) digging. Note the complete lack ofa relation for digging. 15Table 1. Pairwise shown are the Spearman's correlation coefficients p for the correlations between all behavioural measurements. The averages of all 4 measurements from all 4 observations were used for the correlations. p-values marked with () refer to p < 0.001, p-values marked with (") refer to p < 0.01, p- values marked with () refer to p < 0.05 and p-values marked with () refer top < 0. 1
Table 2. Consistency over time; the correlation between the average of the first two and last two observations of the 7 behavioural traits.
Shown are Spearman's S, the rank correlation coefficient p and the p-value.
S p p
compaggro 137.37 07547 0.0011 pred aggro 347.60 0.3793 0.1632 Unique patches 331.89 0.4073 0.1318 Max distance 315.62 0.4364 0.1039
Times away 228.17 05925 0.0199
shelter visits 24537 0.5617 0.0293
Digging 677 18 -0.2092 0.4542
Table 3. Coefficients of the first three principal components showing the contribution of each original variable (the behavioural traits). The coefficients are the eigenvectors scaled such that the sum of squares for each component eouals 1.
Conlpon.nt 1 2 3
camp aggro -0.332 -0.083 -0.690
pred aggro -0.401 0.057 -0.490 uniqu. visits 0.443 -0.196 -0.395
max dIstance 0.454 -0.267 -0.147
#tlm.saway 0.461 -0.160 -0.245
sh•ttir visits -0.308 -0.570 0.180
digging -0.139 -0.728 0 114
Table 4. The standard deviation and eigenvalue (the square rrot of the standard deviation) for each of the 7 components of the PCA and the proportion of the variance it explains. Components 1 to 3 are responsible for almost 90% of all vanance in the 7 behavioural traits.
Component 1 2 3 4 5 7
Standard deviation 1.9278 1.2540 0.9968 0.5308 0.4832 0.4115 0.1817
Elgenvalue 3.7165 1.5724 0.9935 0.2818 0.2335 0.1693 0.0330
Proportion of vanance 0.531 0.225 0.142 0.040 0.033 0.024 0.005 Cumulative proportion 0.531 0.756 0.897 0.938 0.971 0.995 1.000
16
aggr. comp. aggr pred. times away unique visits max dist. digging
U aggressive acts against predator 0.74 *
U times away from territory 0.35 0.57 *
U unique patches visited 0.26 0.55*
0.87
maximum distance from territory 0.46" 0.65 0.78 0.93
U digging acts 0.14 0.13 0.12 0.076 0.11
#timesvisitedtheshelter 0.25 0.41 0.51' 0.35 0.3 0.52
Component correlates
None of the three components differed between the sexes (Welch t test, 1g component:
t=0.6186, df.= 4.9 17, p=0.5637; 2' component t=-0.1739, df.= 7.817, p=0.8664; 3 component: t=O.9286, df. = 11.681, p=0.3719), though the variance In the first component was much larger for the smaller group (males), but not significantly so (F-ratio test; F = 0.2235; df 9 / 4; p = 0.05794).
The amount of feeding (fig. 6), measured as the number of feeding bites taken during an observation, did not correlate with the major principal component, though had a significant
negative correlation with the third component (see table 5), thus an increase in this component (less aggression and patch visiting diversity) was associated with a decrease in feeding of the helpers.
Spearman's rank correlation analyses revealed that the first principal component was significantly correlated with the size of the breeding group the focal helper is affiliated with. In larger groups the chance of finding a more aggressive, less exploring individual was therefore higher (see fig. 7 and table 5).
Density of predators was not a good predictor for any of the three components (Spearman's rank correlation, see table 5). The density of predators was lower at the rim compared to the centre of a colony (see fig. 8a; t test, t=2.645, df. = 17, p=0.017) as was the second principal component (see fig. 8b; Welch t test, t=2.6499, df.=8.161, p=O.02876), but not the other two components (t test, 1g component: t=1.3195, df. = 13, p=O.2098; 3 component: t=O.0113, df=13, p=0.9912). This Implied a relationship between predation and the shelter-visit-digging component. Direct causality was excluded by comparing paired observations of shelter visits with and without presenting a predator In the territory which showed an unsignificant difference (see fig. 8c; paired t test, t=O.8897, df. = 18, p = 0.3854).
Density of territories on a small scale (1m2) correlated significantly with the third component, but not the larger two components (see table 5, fig. 9a,b,c). When counting density on a larger scale (16m2), this density effect on the third component faded out (see table 5, fig. 9d,e,f). At this scale more densely populated areas were associated with higher values of the shelter-visit-digging component, but not significantly so (see table 5).
17
- - — — — — — — — — — — — — — — — — — — —
explorativeno digging'
non aggressive.
•i
0S..5
S.
C)C) N S 0N. I'.
S aggressive..
adigging b
u)ir fl9 55
flp.b(55(fl9 blissgnsC Fig. 6. The correlation between the three principal component values and the number of feeding bites taken during the observation. x-axis shows the number of feeding bites, the y-axis the individuals value of a) behavioural component 1; b) behavioural component 2; c) behavioural component 3. See table 3 for the exact makeup of each component. explorativeno digaing*•non . N -5aggressive.
S•.
• S•. 0-
1. :
0°. : :
aggressiveSdigging•aggressive.
IIIIIIIIIIIIIIIII 4567891011
4567891011
4567891011
a.,
bc Fig. 7. The correlation between the three principal component values (y-axis) and size of the breeding group (x-axis) of the individual. Component 1, personality', decreases with gmup size; so more aggression and less exploration. See for explanation of the y-axis fig. 6 and table 4. 18- — — — — — — — — — — — — — — — — — — —
Table 5. Overview of all Spearman's rank correlations on the three main principal components for every ecolo9ical factor. Shown are Spearman's S, the rank correlation coefficient p and the p-value. SPrinciple corn pponent I p
Pri S
nclple corn p
ponent 2 p
Pr S
inciple corn p
ponent 3 p Feeding bItes439.890.21 450.4427459.910.17870.5239856.26-0.52900.0426 Group size878.95-0.56960.0268'600.13-0.07170.7997601.16-0.07350.7946 Predator density684.00-0.22140.4266338.000.39640.1427620.00-0.10710.7084 Territory density - small scale538.610.03820.8925426.600.23820.3926952.06-070010.0037' TerTitory density - large scale424.760.24150.3859295.530.47230.0755607.08-0.08410.7658 19
0 I0
C0
> U) U)U)-
a) C U)nodigQng digging ba shelter visits witout predatorC Fig. 8. The influence of predator density; a) Predator density is significantly higlier in the centre (c) than on the rim (r) of the N.puicher colony. b) Shelter-visit- dig component (PC 2) is significanfly higher (less visits to the shelter and digging) in the centre of the colony than on the nm. c)Presence of predators does not induce extra visits to the shelter.
S
I • 8• 11:8.
at rimorcentral m the canyatrlmoi-c.nVal th.conyIIIIII 051015202530
— — — — — — — — _
-— — — — — — — — — —
explorative N aggressive explorative N -I
aggressiveS •S S S S
.
S S U S •S III• I 123456 Ioc gtoi densitya S
. . .
SS •
.
S S S S Sno digging N0 p diggina no digging
I
diggingS •S• S •S S • S
• S S
.
S IIIII 123456 csigraup densitybI
S.
S••
S S S S S S S S Snon aggressive 0
g0 00
0 aggressive non aggressive 000
0. I
•. .
S.
S S •S S S IIII 23456 beat gro densityC.
•.•
S S S. .
S• S S• 20304050
6070 local g, densityd203040506070 local cup densityC Fig. 9. The relation between the three different components and density of temtones. a,b,c) Density counted on a small scale (1m2). d, e, f) Density counted on a large scale (1 6m2). Notice how the coffelaon shifts from component 3 on the small scale to component 2 on a largerscale.
IIUIUU 203040506070 local giot* density 20
DISCUSSION
Main findings
Personality is defined a complex of behaviours, all consistent over time and co-varying with each other (Sih eta!. 2004,Reale et a!. 2007); we have demonstrated consistent behaviour of Neolamprologus puicher In the field and that these behaviours were inter-correlated in a behavioural syndrome. Digging excluded, all behaviours made up part of this syndrome, shelter visits as a measure of help Included. Our initial hypothesis that the more active/risk- taking individuals would provide less help, therefore proved to be wrong. Unexpectedly we found a negative correlation between aggression and exploration in the field, in contrast with several other studies (Wilson & McLaughlin 2007,Dingemanse eta!. 2007) including a lab- study on N. pu!cher (SchUrch & Heg 2008). Yet, exploration was negatively correlated with helping behaviour. If exploration is a good predIctor of dispersal, this result would be consistent with the helping investment — dispersal trade-off (Bergmuller eta!. 2005b).
To the contrary of our expectations, the found behavioural syndrome correlated with hardly any life-history or ecological trait; the sexes did not differ in behavioural types so females should be just as likely roamers as males. A genetic analysis revealed has previously revealed that within-colony dispersal does not differ between the sexes either (Stiver et a!.
2007). Feeding as a proxy for growth also showed no relation; difference in growth and a therefore adjusted feeding rate has been proposed as a physiological mechanism inducing consistent variation in behaviour (Stamps 2007). The behavioural types were distributed evenly throughout the colony and independent of local density of predators or competition intensity. If there would be any selection for certain behavioural types on such a local scale, it would no doubt be counteracted by the large within-colony dispersal found in an earlier study (Stiver eta!. 2007).
Our aggression-exploration-helping syndrome did correlate with group size. Individuals from larger groups were more aggressive. Whether this also means more conflict with group members Is unknown; the introduced conspecific used in the tests was always an Individual from outside the colony, so probably a complete unknown individual to our aggressive acting focal. Larger groups are known to be more stable over the years (Heg eta!. 2005) and to have better quality territories (Balshine eta!. 2001); Submissive helpers might have to provide more help to remain tolerated by the dominant breeders and other group members ("pay-to- stay": Balshine-Earn eta!. 1998,Bergmullereta/. 2005b). Alternatively one could consider that the high competition in the large queue does not allow individuals to stay away for to long or loose their position. Would that mean that individuals beyond a certain 'threshold type' will always be evicted and never become breeders? Or do they use alternative strategies to get into a breeding position? A behavioural type could be status-dependent, causing individuals to adjust their behaviour to their breeding prospects. Physiological adjustments in size and growth to status are already known in this species (Heg eta!. 2004).
21
I
I
I
Consistency and inter-correlation
Aggression towards a competitor, number of departures from the territory and number of
I
shelter visits were significantly consistent over time. In a laboratory study on the same species (SchUrch & Heg 2008), aggression towards a competitor was only significant forI
mates, defence against predators only for females and sand carrying and exploration for both sexes. The actual experiments done to acquire these trait quantifications differed from ours and therefore exact comparisons are invalid. What It does show is that results found in one
I
population cannot easily be extrapolated over the whole species, as already shown by Dingemanse et aI.(2007).The three spatiai measurements (number of unique territories visited, maximum
I
distance from home territory and number of times away from the home territory) were strongly correlated with each other. This is not to surprising since the measurements wereI
taken from the same observation and would therefore be intrinsically correlated. But it did show that individuals that leave the home territory more often did not just visit the same close neighbour over and over, but seem to sample the whole colony more intensely.
I
When roaming around, an individual has less time to spend on territory maintenance i.e. digging. One might therefore expect an intrinsic negative relation between the amount of digging performed and the number of times away from the territory, especially since theI
measures were taken from the same observation. Yet, this relation was absent, showing that this intrinsic relation did not influence the correlation.I
The other componentsThe 'helping syndrome' correlated significantly with the location of the territory in the colony.
I
Individuals at the rim of the colony digged more and paid more visits to the shelters. These shelter visits were not induced by predators, indeed the correlation between predator density and the digging component suggests the opposite. Moreover, the comparison between 1 treatments showed no relationship between presence of a predator and the number of sheltervisits.
I
At the rim, the territory density was lower, suggesting lower competition levels. Since the correlation between digging and territory density only showed when this density was measured at a large scale, the competition was more lIkely to be over food (which is
I
consumed in a large area around the territory) than over territorial area. In quieter areas, less food competition arises therefore more time would be available to move sand around. Another explanation for more digging at the rim would be that the substrate tended to be more sandyI
on the outer stretches of the colony, making these territories higher in maintenance. If the 'syndrome' was maintenance-dependent, it would also explain the non-consistency of the behaviour. Perhaps an experimental manipulation of the state of the territory would haveI
given a clearer picture, though standardizing this in the field is a big challenge.I
22I
I
When local density is measured on a small scale, this provides a good estimation of territorial pressure by the neighbours. The second component did not show any relation with this density measure, yet the third component did; in the denser areas (when a territory had many neighbours close by), our focal helpers were more aggressive. With a fierce competition for space, more aggressive conflicts would occur. An individual might therefore have its aggression level increased accordingly. However, this was not compensated by reducing the other behaviours we monitored; most of them even seem to increase (though only slightly).
Of all the other factors did only the number of feeding bites also correlate with the third component. So the Individuals that Increased aggression levels due to stronger space competition ate more. We lacked the exact data to test this hypothesis; a controlled experiment would be necessary.
Measurement considerations
It is Important to realize what one is exactly measuring (Bell 2007); how similar Is our exploration measure to exploration in other studies? Most studies introduce the caught Individual to a completely new enclosed environment (Dingemanse eta!. 2003,Wilson &
McLaughlin 2007,Dochtermann & Jenkins 2007). In that sense our spatial measure was not 'exploration' since the environment Is, presumably, known to our focal. But the environment is ever-changing, so to compare It just to a general 'activity level' in the tank/cage does not do justice either (which was also expected to be positively correlated to the boldness syndrome).
Fleeing behaviour is a possibility though these roaming measurements were taken when no predator or competitor was in the tube at the territory. It resembles the previously found keeping track of the surroundings, In search of good dispersal opportunities (Bergmuller et a!.
2005a). And the less aggressive types would do this more and further. The Chitty-Krebs hypothesis already predicted docile individuals to disperse more, forced by their more
aggressive counterparts (Krebs 1978). The relation between exploration and dispersal, though found in other studies (Dingemanse eta!. 2003), could not be shown here.
The number of unique territories visited and the maximum distance away from the territory were in some cases underestimations of the true value; during an observation the focal would be lost from sight when it strayed too far away from the observer. For some Individuals this might have made quite a difference for their measured values.
What we exactly measured Is also a valid question for the aggression measurements.
Even though the Lepidiolarnpro!ogus elongatus might have been a personal threat, the predator was introduced to the focal at the breeding territory, making it hard to exclude, or rather likely an altruistic act of the focal to attack the L. elongatus. Balshine eta!. (1998) considered these attacks as helping effort. Attacking the similar sized conspecific can more likely be considered a true selfish act: chasing away possible extra helpers because they form a threat to the focal's personal dominance position. Still, both our aggression measurements
23
were taken at the territory. Individuals who cared less about the fate of the territory might therefore have reacted less severe.
As a measure of true cooperation shelter visits and digging were used. These
measures turned out to be ambivalent in the field. The altruistic level of visiting a shelter was hard to asses when it is not known in which shelter the brood chamber resides; further more, N.pulcher helpers are known to cannibalize the brood (SchUrch, personal communication) which was impossibie to determine in the field. The function of the shelter visits for our focals (brood care, brood cannibalism, sheltering) therefore remained inconclusive. Digging is a form of territory maintenance, but if only performed on the focal's own shelter, the altruistic level of it remains questionable. Moreover, individuals that invest less in these traditional measures of cooperation (i.e. brood care) are known to compensate by other means (e.g. predator
mobbing; Arnold eta!. 2005).
Digging was a completely Inconsistent trait. In concordance it did not show up in the component of the behavioural type. Instead, a separate component needed to be constructed to explain this variation in our population. Naturally, shelter visiting loaded into this
component as well; one has to enter the shelter in order to remove sand from it. Since the shelter visits also loaded well into the first component, one Is tempted to say that these visits differ in function from the shelter visits in the second component. Truth is that there was no data to support this. Labelling this component as a 'helping syndrome' would therefore be premature though not prodigious (see: Balshine-Earn eta!. 1998).
Discerning helpful behaviours from selfish acts could ultimately be considered impossible.
Behaviours we assumed to be helpful could never be proven as such, though admittedly, this is partly semantics. Aggression against predators, initially used as a measure for the helper's behavioural type, could also be a good estimator for the level of help provided. One could argue that using the same measure as both an estimator of aggression and an estimator of help causes a trivial intrinsic perfect correlation. Yet the helper will be confronted with this in its natural life as well. Better designed behavioural tests, away from the helper's territory should in these cases separate the purely aggressive from the altruistic, if this distinction exists at all. For we have just demonstrated these correlations among behaviours. The long- term consequences in the natural world remain to be identified.
24
1
1
I
ACKNOWLEDGEMENTS
The research for this master's project involved a lot of practical work and I am in debt of many
I
people for making this possible. First of all I would like to thank Roger Schurch, my 'field assistant' yet a great supervisor for trusting me. I am very grateful that we got along so wellI
during the field expedition. For this I am also very thankful to Alex Kotrschal, Martin Jiskra and Oliver Otti. Oliver also helped In the fieldwork and in writing part of this thesis.
A sincere natasha 'mkwal to Bonface and his family for taking care us during our stay
I
and lending whatever we asked for. A special thanks for Celestine who prepared divine meals for us daily. Rueben Shapola and Danny Sinyinza from the local fisheries department have been invaluable help to us In various matters; from repairing compressors to arranging gasoline to helping out with visas. Last to thank from Zambia Is Phil Nielsen for providing us diving essentials and —excursions.I
I am In gratitude of DIk Heg for allowing me to do this project and helping me out In Bern on the data analyses. I thank Theunis Piersma for keeping an eye on my progress. And I thank my parents for keeping an eye on me my whole life and supporting me this venture.
I
Financially this study has been made possible by the Groninger Unlversiteitsfonds, the Marco Polo fonds and grant 3100A0-108473 of the Swiss National Science Foundation to DikI
I I
I I I
I I
I
251
I
I
I
I
REFERENCES
I
Arnold, K. E., I. P. F. Owens, and A. W. Goldizen. 2005. DivIsion of labour within cooperatively breeding groups. Behaviour 142:1577-1590.I
Baishine, S., B. Leach, F. Neat, H. Reid, M. Taborsky, and N. Werner. 2001. Correlates of group size in a cooperatively breeding clchlid fish (Neolamprologus pulcher).I
Behavioral Ecology and Soclobiology 50: 134-140.
Balshine-Earn, S., F. C. Neat, H. Reid, and M. Taborsky. 1998. Paying to stay or paying to
I
breed? Field evidence for direct benefits of helping behavior in a cooperatively breeding fish. Behavioral Ecology 9:432-438.
I
Bell, A. M. 2007. Future directions in behavioural syndromes research. Proceedings of the Royal Society B-Biological Sciences 274:755-761.
I
Bell, A. M., and A. Sih. 2007. Exposure to predation generates personality in threespined sticklebacks (Gasterosteus aculeatus). Ecology Letters 10:828-834.
I
Bergmuller, R., D. Heg, K. Peer, and M. Taborsky. 2005a. Extended safe havens and between- group dispersal of helpers in a cooperatively breeding clchlid. Behaviour 142:1643- 1667.
Bergmuller, R., D. Heg, and M. Taborsky. 2005b. Helpers in a cooperatively breeding cichlid stay and pay or disperse and breed, depending on ecological constraints. Proceedings
I
of the Royal Society B-Biological Sciences 272:325-331.Cahan, S. H., D. T. Blumstein, L. Sundstrom, J. Liebig, and A. Griffin. 2002. Social trajectories
I
and the evolution of sociai behavior. Oikos 96:206-2 16.Chadwick Johnson, J., and A. Sih. 2005. Precopulatory sexual cannibalism in fishing spiders
I
(Dolomedes triton): a role for behavioral syndromes. Behavioral Ecology and Sociobiology 58:390-396.I
Clutton-Brock, 1. 2002. Behavioral ecology - Breeding together: Kin selection and mutualism in cooperative vertebrates. Science 296:69-72.Dali, S. R. X., A. I. Houston, and J. M. McNamara. 2004. The behavioural ecology of personality: consistent individual differences from an adaptive perspective. Ecology Letters 7:734-739.
1
261
I
Dierkes, P., D. Heg, M. Taborsky, E. Skubic, and R. Achmann. 2005. Genetic relatedness in groups Is sex-specific and declines with age of helpers in a cooperatively breeding cichlld. Ecology Letters 8:968-975.
Dingemanse, N. J., C. Both, A. J. van Noordwijk, A. L. Rutten, and P.). Drent. 2003. Natal dispersal and personalities in great tits (Parus major). Proceedings of the Royal Society of London Series B-Biological Sciences 270:741-747.
Dingemanse, N. )., J. Wright, A.). N. Kazem, D. K. Thomas, R. Hickling, and N. Dawnay.
2007. Behavioural syndromes differ predictably between 12 populations of three- spined stickleback. Journal of Animal Ecology 76:1128-1138.
Dochtermann, N. A., and S. H. Jenkins. 2007. Behavioural syndromes In Merriam's kangaroo rats (Dipodomys merriami): a test of competing hypotheses. Proceedings of the Royal Society B- Biological Sciences 274:2343-2349.
Godin,). G. 3., and L. A. Dugatkin. 1996. Female mating preference for bold males in the guppy, Poecilia reticulata. Proceedings of the National Academy of Sciences of the United States of America 93:10262-10267.
Hamilton, W. D. 1964. Genetical Evolution of Social Behaviour I. Journal of Theoretical Biology 7:1-16.
Hatchwell, B. J., and 3. Komdeur. 2000. Ecological constraints, life history traits and the evolution of cooperative breeding. Animal Behaviour 59:1079-1086.
Heg, 0., and Z. Bachar. 2006. Cooperative breeding in the lake tanganyika cichlid Julidochromis ornatus. Environmental Biology of Fishes 76:265-281.
Heg, D., N. Bender, and I. Hamilton. 2004. Strategic growth decisions in helper cichlids.
Proceedings of the Royal Society of London Series B-Biological Sciences 27 i:S505- S508.
Heg, D., L. Brouwer, Z. Bachar, and M. Taborsky. 2005. Large group size yields group stability in the cooperatively breeding cichlld Neolamprologus pulcher. Behaviour 142:1615-
1641.
Huntingford, F. A. 1976. Relationship Between Anti-Predator Behavior and Aggression Among Conspecifics in 3-Spined Stickleback, Gasterosteus-Aculeatus. Animal Behaviour 24:245-260.
27
Komdeur, J. 2006. Variation in individual Investment strategies among social animals.
Ethology 112:729-747.
Komdeur, J., and P. Edelaar. 2001. Evidence that helping at the nest does not result In territory inheritance In the Seychelles warbler. Proceedings of the Royal Society of London Series B-Biological Sciences 268:2007-2012.
Krebs, C.). 1978. Review of the Chitty Hypothesis of Population Regulation. Canadian Journal of Zoology-Revue Canadienne de Zoologie 56:2463-2480.
Moretz, J. A., E. P. Martins, and B. D. Robison. 2007. Behavioral syndromes and the evolution of correlated behavior in zebrafish. Behavioral Ecology 18:556-562.
Pen,!., and F..). Weissing. 2000. Towards a unified theory of cooperative breeding: the role of ecology and life history re-examined. Proceedings of the Royal Society of London Series B-Biological Sciences 267:2411-2418.
Price, T., and T. Langen. 1992. EvolutIon of Correlated Characters. Trends In Ecology &
Evolution 7:307-310.
Reale, D., S. M. Reader, D. Sol, P. T. McDougall, and N.). Dingemanse. 2007. Integrating animal temperament within ecology and evolution. Biological Reviews 82:291-318.
SchUrch R. & Heg D. Life history, behavioural type and cooperation in the highly social clchlid Neolamprologus pulcher. submitted.
Sih, A., A. Bell, and J. C. Johnson. 2004. Behavioral syndromes: an ecological and evolutionary overview. Trends in Ecology & Evolution 19:372-378.
Smith, B. R., and D. T. Blumstein. 2008. Fitness consequences of personality: a meta- analysis. Behavioral Ecology 19:448-455.
Stamps,). A. 2007. Growth-mortality tradeoffs and 'personality traits' in animals. Ecology Letters 10:355-363.
Stapley, J., and). S. Keogh. 2005. BehavIoral syndromes influence mating systems: floater pairs of a lizard have heavier offspring. Behavioral Ecology 16:514-520.
Stiver, K. A., J. K. Desjardins, J. L. Fitzpatrick, B. Neff,). S. Quinn, and S. Balshine. 2007.
Evidence for size and sex-specific dispersal In a cooperatively breeding cichlld fish.
Molecular Ecology 16:2974-2984.
28
I
I
I
I
Werner, N. Y., S. Balshine, B. Leach, and A. Lotem. 2003. Helping opportunities and space segregation in cooperatively breeding cichllds. Behavioral Ecology 14:749-756.I
Wilson, A. D. M., and R. L. McLaughlin. 2007. Behavioural syndromes in brook charr, Salvelinus fontinalis: prey-search In the field corresponds with space use In novelI
laboratory situations. Animal Behaviour 74:689-698.
Wolf, M., G. S. van Doom, 0. Leimar, and F. J. Weissing. 2007. Life-history trade-offs favour the evolution of animal personalities. Nature 447:581-584.
I
I I
I I
I I
I I I I
I
29I
I
APPENDIX A
Table I. List of species found in the habitat of Neo!ampro!ogus puicher. Type refers to the type used in the analyses for predator density and density of temtones. Predation in this case can mean anything from a brood parasite to a true piscivore. Note that a single species can be both a predator and space competitor.
Species name Type
Lepidiolamprologus elongatus Lepidiolamprologus lemaini
Predator! Space competitor Predator Lepidio!amprologus attenuatus Predator
Mastacembelus sp. Predator
Altolampro!ogus calvus Predator
Neo!ampro!ogus sex fasciatus Predator
Gnafhochmmis pfeffen Predator
Synodontis sp. Predator
Penssodus microlepis Predator
Juiidochromis omatus Space competitor
Lamprologus callipterus Space competitor Neolamprologus modestus Space competitor Neo!amprologus tetracanthus Space competitor Te!matochromis tempora!is Space competitor Telmatochromis vittatus Space competitor Neolamprologus caudopunctatus Space competitor
Neolamprologus pulcher Focal Species/
Space competitor
30
