# High cost enhances cooperation through the interplay between evolution and self-organisation

- Enrico Sandro Colizzi
^{1}Email author and - Paulien Hogeweg
^{1}

**16**:31

https://doi.org/10.1186/s12862-016-0600-9

© Colizzi and Hogeweg. 2016

**Received: **6 October 2015

**Accepted: **22 January 2016

**Published: **1 February 2016

## Abstract

### Background

Cooperation is ubiquitous in biological systems, yet its evolution is a long lasting evolutionary problem. A general and intuitive result from theoretical models of cooperative behaviour is that cooperation decreases when its costs are higher, because selfish individuals gain selective advantage.

### Results

Contrary to this intuition, we show that cooperation can increase with higher costs. We analyse a minimal model where individuals live on a lattice and evolve the degree of cooperation. We find that a feedback establishes between the evolutionary dynamics of public good production and the spatial self-organisation of the population. The evolutionary dynamics lead to the speciation of a cooperative and a selfish lineage. The ensuing spatial self-organisation automatically diversifies the selection pressure on the two lineages. This enables selfish individuals to successfully invade cooperators at the expenses of their autonomous replication, and cooperators to increase public good production while expanding in the empty space left behind by cheaters. We show that this emergent feedback leads to higher degrees of cooperation when costs are higher.

### Conclusions

An emergent feedback between evolution and self-organisation leads to high degrees of cooperation at high costs, under simple and general conditions. We propose this as a general explanation for the evolution of cooperative behaviours under seemingly prohibitive conditions.

### Keywords

Evolution of cooperation Spatial model Viscous populations Travelling waves Multi-level evolution## Background

When cooperation is costly to the individual but its benefits are equally shared in a group, one would expect progressively more selfish behaviour to be selected. This indeed happens when interactions between cooperative and selfish individuals are random. Instead, cooperation can be selected in a population when the interactions among cooperative and selfish individuals are structured, be it genetically, spatially or socially [1–6]. Population structure favours cooperation when it allows for cooperators to be in contact with each other more frequently than with selfish individuals [7].

The result is a (locally or globally) stable equilibrium configuration in which cooperators persist indefinitely, and selfish individuals may co-exist. For instance, this can be due to spatial clustering of cooperators [8], or to an inherent structure of the interaction network [9]. Then, the conditions for selfish individuals invading and overtaking a group of cooperators represent the limit to the stability of the solutions found. In general, these conditions state that higher costs of cooperation increase the selective advantage of selfish individuals.

While spatial structure alone can favour cooperators due to population viscosity, a growing body of experimental and theoretical work indicates that self-organised spatial patterns may have profound and complex effects on cooperative interactions, due to emergent heterogeneities in the local distributions and densities of cooperators and selfish individuals [10–19]. For instance, as a population of cooperators invade empty space, its expansion front can be enriched in altruistic individuals, while selfish individuals lag behind [13, 15, 18]. Alternatively, mutualistic interactions that are favoured in a resident population can be automatically broken on such expansion front [14]. Furthermore, spatial self-organisation can sort cooperative strains from selfish ones [12, 20–24], thus limiting the spread of the latter.

The customary approach to study the stability of cooperation (under a specific set of assumptions) consists of fixing the strategies of the interacting cooperative and selfish individuals, and analyse the population dynamics of the system (e.g. whether cooperators and defectors coexist, or if one lineage outcompetes the other). Because such pre-determined strategies do not mutate over time, their evolutionary stability remains unexplored.

Although exceptions to this approach exist (e.g. [8, 22, 25, 26]), little is known about what strategies evolve (by mutation and selection) and how they feedback on the spatial self-organisation of a population, even though it is clear that spatial self-organisation affects the population dynamics of cooperative traits (see examples above). Here, we seek to study this feed-back between evolution and self-organisation with a minimal model where individuals can evolve the degree of cooperation in a spatially extended system.

We model the cooperative trait in terms of public good production (inspired by social dynamics in microbes [27, 28]), and we let the amount of public good produced mutate in a continuous fashion. Thus, we can study the long term evolutionary dynamics of cooperation without preconceiving the extent to which individuals cooperate or defect. As we will show, selfish individuals that produce zero public good evolve readily at higher costs, and quickly invade cooperators. Rather than leading to global extinction, this enables cooperators to thrive and selects for a higher degree of public good production over evolutionary time scales.

## Results

The model is a straightforward implementation of a population in which individuals replicate depending on the amount of public good produced in their close neighbourhood.

### High cost leads to the evolution of larger public good production

We set the benefits per unit of public good b=10, and we let the spatial self-organisation and the evolutionary dynamics unfold under different costs c.

Strikingly, further increasing costs leads to an increase in cooperation (c>3, Fig. 2, see Additional file 1: Section 1 for the full snapshots of the system). The distribution of public good production is bimodal at evolutionary steady state, with most of the population having higher rates of public good production and a minority producing almost no public good at all.

### The long term evolutionary dynamics of cooperation at high cost

New waves are “born” from the collisions of older waves. As cooperators on the front evolve to larger public good production, waves become larger (compare Fig. 3 e, f and g, and Additional file 1: Section 2 for the full snapshots). The formation of spatial patterns allows the populations in the system to persist indefinitely despite selfish individuals continuously invading cooperators (Fig. 3 g and Additional file 1: Section 6), provided that the lattice is much larger than the spatial patterns (see Additional file 1: Section 3).

### Spatial population dynamics of cooperators and selfish individuals

Spatial self-organisation drives the evolution of cooperation in the system. When spatial patterns are destroyed, e.g. by mixing, only selfish individuals are selected and public good production decreases, leading to global extinction (Additional file 1: Section 4), in accordance with the result that random interactions favour selfish behaviour.

The replication rate of selfish individuals invading a population of cooperators is higher when the cost of public good production is higher, and it is insensitive to how much public good is produced (Fig 5 a, blue). Clearly, when two strains compete in the back of a wave, the winner is the more selfish one (Fig 5 b).

The picture emerging from these experiments is that different selection pressures operate depending on the spatial context: a population expanding into empty space (the wave front) is selected for higher degrees of cooperation (in agreement with [12, 13, 15, 18]), competition in the back (behind the wave front) selects for more selfishness. Importantly, even though costs, benefits and fitness function are the same, spatial pattern formation automatically segregates these two opposing evolutionary pressures to spatially different contexts so that they do not balance each other: hence the evolution of a cooperative and a selfish lineage.

In the full system, cooperators and selfish individuals are ecologically and evolutionarily interdependent (Additional file 1: Section 5). The two lineages establish an evolutionary feedback mediated by their spatial organisation. Empty space is generated by selfish individuals after invading a population of cooperators. Therefore, the condition for increasing public good production, i.e. the availability of continuous empty space, is mediated by the invasion dynamics of the selfish lineage. With higher costs, selfish individuals propagate faster, and space is left empty at a higher rate. The larger the empty space, the more cooperators can increase public good production. This evolutionary feedback reaches an evolutionary steady state because highly producing cooperators reduce the empty space faster upon faster expansions (Fig 5 a). We checked the long term stability of the steady state (Additional file 1: Section 6).

Altogether, cooperation evolves to a higher degree for higher costs due to an emergent feedback between self-organised interaction structures (the spatial patterns) and the evolution of the individuals composing them.

### Robustness to parameter change

Our results are robust when death rates, movement rates and benefits are changed, provided that benefits-to-costs ratio is maintained (Additional file 1: Section 7). At lower costs, individuals directly benefiting themselves with their own public good are sometimes dubbed *weak* altruists [7, 30] (in our case, for *c*<*b*/(*k*+1), with *k*+1=9 the connectivity of the lattice including self), whereas they are considered *strong* altruists when their public good is only shared among others (in the context of game theory, these situations are called, respectively, snowdrift game and prisoner’s dilemma [25]). In our spatial model, we observed no qualitative difference in the evolutionary dynamics when individuals did or did not benefit from their own public good (Additional file 1: Section 8). This could be expected because individuals’ own payoffs at high costs are negative in both models. Indeed, in both cases the evolving populations underwent speciation of a selfish and a cooperative lineage.

Weak or strong altruism do make a difference in the corresponding well-mixed systems, where strong altruists go extinct at lower costs, while weak altruists maximise public good production (Additional file 1: Section 9).

## Discussion

It is known that during population range expansion, cooperation can be promoted on the front of the expansion range [12, 13, 15, 18]. One could argue that in these models cooperation could evolve only as long as empty space is available, and should eventually be out-competed globally by selfish strategies when the invasion dynamics reach an end. Here we have shown that selfish individuals provide the empty space to allow continuous expansion within a limited area.

Generally, stable solutions to the problem of cooperation are based on the condition that at equilibrium selfish individuals do not locally invade a population of cooperators. For instance, it is well known that cooperation can be stabilised in spatially extended systems, as cooperators cluster and segregate selfish individuals to the edges of those clusters [8, 19, 31]. A side effect of these solutions is that higher costs undermines the stability of such clusters. In the parameter region where travelling waves do not form (Fig. 2, for c≤3) our work is in agreement with those results in that there exists an inverse relationship between costs and cooperation (e.g. [1, 2, 6]), and in particular with the heuristics [9] that cooperation evolves when the benefit-to-cost ratio is larger than the connectivity of the lattice (b/c>*k*, while it becomes progressively more unlikely when b/c approaches *k*).

However, we have shown that a novel class of solutions exists at high costs, where large degrees of cooperation are maintained in a locally out-of-equilibrium fashion, with selfish individuals always successfully invading cooperators and setting the stage for the evolutionary increase and the global stability of cooperation. We conclude that spatial self-organisation can reverse the relationship between costs and cooperation, thus extending the evolutionary viability of cooperation to higher costs.

Our results rest on two assumptions: population size can vary and some degree of cooperation is necessary for reproduction. Variable population size is obviously realistic, even though it is not often included in evolutionary models of cooperation. Although the assumption of necessary cooperation is not always met, it is reasonable in several cases. Examples in microbiology include, cooperative protection or cooperative virulence in bacterial infections [32, 33]; invertase production in yeasts while growing on sucrose [34]; siderophore production in iron-limited environments [35–37], cooperative secretion of digestive enzyme in microbial hunting [27, 38]. Outside the microbial world, situations where our model may apply are e.g. dangerous behaviours in cooperative nest defence [39], and replication *in trans* in prebiotic evolution [40, 41].

Two recent studies have come to conclusions that at first sight are similar to ours [17, 19]. By making the assumption, as we do, that the lack of cooperation leads to death, they observe (quasi) static spatial patterns in which cooperation is maintained because despite relatively high costs, clusters of cooperators cannot be invaded by selfish individuals.

In contrast, cheaters can always invade in our system, and do so faster when costs are higher. This shows that costs are qualitatively higher in our model. Cooperation is maintained despite, and due to the evolution of true cheaters (in the sense of [42]). Furthermore, we show that the amount of public good produced by an individual increases in evolutionary times, whereas the evolutionary stability of the solutions in [17, 19] is left unexplored, and only the long population dynamical transient is analysed.

More in general, the importance of spatial self-organisation for understanding the population dynamics of cooperators and defectors has been highlighted both from a theoretical [10] and from an experimental point of view [43]. Here we make a similar point, but with an evolutionary twist: in our case selfish individuals are not merely a burden to cooperators; instead, their emergence as a separate lineage is necessary for the evolution of high degrees of public good production because they generate the spatial conditions in which cooperators thrive and evolve.

Allowing mutations to change public good production in a continuous range resulted in the evolution of two separate strains, a selfish and a cooperative one. The evolution of stable heterogeneity in a population has been observed before in models of cooperation [22, 25, 44]. Here, besides stressing that the evolution of two lineages from a single ancestral one might be a rather general feature of models with variable investments (as very simple assumptions were needed, in contrast to [25]), we make the case that true cheating behaviour (*sensu* [42]) is actually functional and beneficial to the long term evolution of cooperation.

## Conclusions

In conclusion, besides extending the theoretical limits of cooperation, our results broaden the search image of cooperative behaviour in nature by suggesting that there need not be a strict trade-off between costs and benefits; rather, a wider view of the self-organised eco-evolutionary processes must be taken into account to understand the occurrence of costly cooperation.

## Methods

### General system

Our system is an individual-based, Monte Carlo simulation run on a square lattice with connectivity *k*=8 and toroidal boundary conditions. The nodes of the lattice can be empty or occupied by at most one individual. Individuals produce public good with rate *p* per time step (alternatively, *p* can be considered the degree of altruism of an individual). An individual *i* produces *p*
_{
i
} public good per time step (0≤*p*
_{
i
}≤10), which is divided equally among neighbouring nodes and self, each receiving *p*
_{
i
}/9 public good. All *n* neighbours, in turn, share a fraction of the public good they produce *p*
_{{1,2,…,n}}/9 with individual *i*. The benefit from the public good received from each neighbour and from self is \(B_{i} = \mathrm {b} \left (\frac {p_{i}}{9} + \frac {1}{9} \sum \limits _{j=1}^{n} p_{j} \right)\), where b is the benefit per unit of public good. Individuals pay a cost proportional to the public good they produce C_{
i
}=c*p*
_{
i
}. Public good is not accumulated over multiple time steps. The fitness of an individual is the difference between benefits and costs: *f*
_{
i
}=B_{
i
}−C_{
i
} (set to zero if costs exceed benefits).

Each Monte Carlo step, all nodes are updated in random order (although synchronous updating rules do not affect results). If a node is empty, the individuals in its neighbourhood (if any) compete for replication. Assume an individual *i* is competing with *m* other individuals, and let us name *f*
_{tot} the sum of the fitness of all individuals competing for the same empty node, then \(f_{\text {tot}} = f_{i} + \sum \limits _{k=1}^{m} f_{k}\). Individual *i* is chosen for reproduction over its competitors with probability \(\mathrm {P}(i~\text {replicates}) = \frac {f_{i}}{ f_{\text {tot}}} \left (1 - e^{-f_{\text {tot}}} \right)\). The term in parenthesis is the probability that at least an individual replicates, which models the assumption replication should be more frequent in a neighbourhood where there is more public good, and conversely it should be rare if little public good is produced. Notice that this term does not affect death.

Upon successful replication, mutations may happen with probability *μ* and affect *p* by adding a small random number drawn with uniform probability from the interval [−*δ*/2,*δ*/2]. If a node is not empty, with probability *k*
_{move} its content is swapped with that of a randomly chosen adjacent node. Moreover, every non empty node can turn to empty with probability *k*
_{death}. See Fig. 1 for a cartoon of the model and the caption of Fig. 2 for the actual values of the parameters. The algorithm is implemented using the CASH libraries [45].

### Invasion dynamics of cooperators and selfish individuals (Fig. 5)

We modified the system described above as follows: 1) we shaped the lattice into a narrow strip of arbitrary length; 2) we changed the boundary conditions to no-flux, and in particular we removed individuals when they moved or replicated into a boundary node of the lattice; 3) we set mutation rates to zero to better focus on spatial population dynamics. The rules for the local dynamics remained the same as above. We initialised all populations on one side of the lattice and waited until they reached the other side. For cooperators this meant that they invaded empty space, whereas selfish individuals invaded a resident homogeneous population of cooperators. In all cases, the number of Monte Carlo time steps it took for the first individuals to arrive to the other side of the lattice (generation by generation) was recorded. The invasion rate plotted in Fig. 5 was calculated as the length of the space invaded divided by the time it took for the population to invade it.

## Availability of data

The code used to run the simulations is available at http://bioinformatics.bio.uu.nl/enrico.

## Declarations

### Acknowledgements

ESC thanks Nobuto Takeuchi, Renske M. A. Vroomans for valuable discussions, and R. Hermsen for suggestions on the manuscript.

**Open Access** This article is distributed under the terms of the Creative Commons Attribution 4.0 International License(http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

## Authors’ Affiliations

## References

- Hamilton WD. The genetical evolution of social behaviour. i. J Theor Biol. 1964; 7(1):1–16.View ArticlePubMedGoogle Scholar
- Wilson DS. A theory of group selection. Proc Natl Acad Sci. 1975; 72(1):143–6.PubMed CentralView ArticlePubMedGoogle Scholar
- Nowak MA, May RM. Evolutionary games and spatial chaos. Nature. 1992; 359(6398):826–9.View ArticleGoogle Scholar
- Mitteldorf J, Wilson DS. Population viscosity and the evolution of altruism. J Theor Biol. 2000; 204(4):481–96.View ArticlePubMedGoogle Scholar
- Santos FC, Pacheco JM. Scale-free networks provide a unifying framework for the emergence of cooperation. Phys Rev Lett. 2005; 95(9):098104.View ArticlePubMedGoogle Scholar
- Van Baalen M, Rand DA. The unit of selection in viscous populations and the evolution of altruism. J Theor Biol. 1998; 193(4):631–48.View ArticlePubMedGoogle Scholar
- Fletcher JA, Doebeli M. A simple and general explanation for the evolution of altruism. Proc R Soc B Biol Sci. 2009; 276(1654):13–19.View ArticleGoogle Scholar
- Killingback T, Doebeli M, Knowlton N. Variable investment, the continuous prisoner’s dilemma, and the origin of cooperation. Proc R Soc Lond Ser B Biol Sci. 1999; 266(1430):1723–8.View ArticleGoogle Scholar
- Ohtsuki H, Hauert C, Lieberman E, Nowak MA. A simple rule for the evolution of cooperation on graphs and social networks. Nature. 2006; 441(7092):502–5.PubMed CentralView ArticlePubMedGoogle Scholar
- Wakano JY, Nowak MA, Hauert C. Spatial dynamics of ecological public goods. Proc Natl Acad Sci. 2009; 106(19):7910–4.PubMed CentralView ArticlePubMedGoogle Scholar
- Nahum JR, Harding BN, Kerr B. Evolution of restraint in a structured rock–paper–scissors community. Proc Natl Acad Sci. 2011; 108(Supplement 2):10831–8.PubMed CentralView ArticlePubMedGoogle Scholar
- Momeni B, Waite AJ, Shou W. Spatial self-organization favors heterotypic cooperation over cheating. Elife. 2013; 2:00960.Google Scholar
- Van Dyken JD, Müller MJ, Mack KM, Desai MM. Spatial population expansion promotes the evolution of cooperation in an experimental prisoner’s dilemma. Curr Biol. 2013; 23(10):919–23.PubMed CentralView ArticlePubMedGoogle Scholar
- Müller MJ, Neugeboren BI, Nelson DR, Murray AW. Genetic drift opposes mutualism during spatial population expansion. Proc Natl Acad Sci. 2014; 111(3):1037–42.PubMed CentralView ArticlePubMedGoogle Scholar
- Datta MS, Korolev KS, Cvijovic I, Dudley C, Gore J. Range expansion promotes cooperation in an experimental microbial metapopulation. Proc Natl Acad Sci. 2013; 110(18):7354–9.PubMed CentralView ArticlePubMedGoogle Scholar
- Suweis S, Simini F, Banavar JR, Maritan A. Emergence of structural and dynamical properties of ecological mutualistic networks. Nature. 2013; 500(7463):449–52.View ArticlePubMedGoogle Scholar
- Szolnoki A, Antonioni A, Tomassini M, Perc M. Binary birth-death dynamics and the expansion of cooperation by means of self-organized growth. EPL (Europhysics Letters). 2014; 105(4):48001.View ArticleGoogle Scholar
- Korolev KS. The fate of cooperation during range expansions. PLoS Comput Biol. 2013; 9(3):1002994.View ArticleGoogle Scholar
- Smaldino PE, Schank JC, McElreath R. Increased costs of cooperation help cooperators in the long run. Am Nat. 2013; 181(4):451–63.View ArticlePubMedGoogle Scholar
- Boerlijst MC, Hogeweg P. Spiral wave structure in pre-biotic evolution: hypercycles stable against parasites. Physica D: Nonlinear Phenomena. 1991; 48(1):17–28.View ArticleGoogle Scholar
- Hogeweg P. Multilevel evolution: replicators and the evolution of diversity. Physica D: Nonlinear Phenomena. 1994; 75(1):275–91.View ArticleGoogle Scholar
- Szabó P, Scheuring I, Czárán T, Szathmáry E. In silico simulations reveal that replicators with limited dispersal evolve towards higher efficiency and fidelity. Nature. 2002; 420(6913):340–3.View ArticlePubMedGoogle Scholar
- Hogeweg P, Takeuchi N. Multilevel selection in models of prebiotic evolution: compartments and spatial self-organization. Orig Life Evol Biosph. 2003; 33(4-5):375–403.View ArticlePubMedGoogle Scholar
- Takeuchi N, Hogeweg P. Multilevel selection in models of prebiotic evolution ii: a direct comparison of compartmentalization and spatial self-organization. PLoS Comput Biol. 2009; 5(10):1000542.View ArticleGoogle Scholar
- Doebeli M, Hauert C, Killingback T. The evolutionary origin of cooperators and defectors. Science. 2004; 306(5697):859–62.View ArticlePubMedGoogle Scholar
- Killingback T, Bieri J, Flatt T. Evolution in group-structured populations can resolve the tragedy of the commons. Proc R Soc Lond B Biol Sci. 2006; 273(1593):1477–81.View ArticleGoogle Scholar
- Crespi BJ. The evolution of social behavior in microorganisms. Trends Ecol Evol. 2001; 16(4):178–83.View ArticlePubMedGoogle Scholar
- West SA, Diggle SP, Buckling A, Gardner A, Griffin AS. The social lives of microbes. Ann Rev Ecol Evol Syst. 2007; 38:53–77.View ArticleGoogle Scholar
- Colizzi ES, Hogeweg P. Videos of the evolutionary dynamics for high costs, and of travelling waves. 2015. http://bioinformatics.bio.uu.nl/enrico. Accessed 29 Jan 2016.
- Wilson DS. Structured demes and trait-group variation. Am Nat. 1979; 113:606–10.View ArticleGoogle Scholar
- Doebeli M, Hauert C. Models of cooperation based on the prisoner’s dilemma and the snowdrift game. Ecol Lett. 2005; 8(7):748–66.View ArticleGoogle Scholar
- Stewart PS, Costerton JW. Antibiotic resistance of bacteria in biofilms. The lancet. 2001; 358(9276):135–8.View ArticleGoogle Scholar
- Diggle SP, Griffin AS, Campbell GS, West SA. Cooperation and conflict in quorum-sensing bacterial populations. Nature. 2007; 450(7168):411–4.View ArticlePubMedGoogle Scholar
- Carlson M, Botstein D. Two differentially regulated mrnas with different 5’ ends encode secreted and intracellular forms of yeast invertase. Cell. 1982; 28(1):145–54.View ArticlePubMedGoogle Scholar
- Miethke M, Marahiel MA. Siderophore-based iron acquisition and pathogen control. Microbiol Mol Biol Rev. 2007; 71(3):413–51.PubMed CentralView ArticlePubMedGoogle Scholar
- Schrettl M, Bignell E, Kragl C, Joechl C, Rogers T, Arst HN, et al.Siderophore biosynthesis but not reductive iron assimilation is essential for aspergillus fumigatus virulence. J Exp Med. 2004; 200(9):1213–9.PubMed CentralView ArticlePubMedGoogle Scholar
- Luján AM, Gómez P, Buckling A. Siderophore cooperation of the bacterium pseudomonas fluorescens in soil. Biol Lett. 2015; 11(2):20140934.PubMed CentralView ArticlePubMedGoogle Scholar
- Konovalova A, Petters T, Søgaard-Andersen L. Extracellular biology of myxococcus xanthus. FEMS Microbiol Rev. 2010; 34(2):89–106.View ArticlePubMedGoogle Scholar
- Olendorf R, Getty T, Scribner K. Cooperative nest defence in red–winged blackbirds: reciprocal altruism, kinship or by–product mutualism?Proc R Soc Lond B Biol Sci. 2004; 271(1535):177–82.View ArticleGoogle Scholar
- Takeuchi N, Hogeweg P. Evolutionary dynamics of rna-like replicator systems: a bioinformatic approach to the origin of life. Phys Life Rev. 2012; 9(3):219–63.PubMed CentralView ArticlePubMedGoogle Scholar
- Higgs PG, Lehman N. The rna world: molecular cooperation at the origins of life. Nat Rev Genet. 2014; 16:7–17.View ArticlePubMedGoogle Scholar
- Jones EI, Afkhami ME, Akçay E, Bronstein JL, Bshary R, Frederickson ME, et al.Cheaters must prosper: reconciling theoretical and empirical perspectives on cheating in mutualism. Ecol Lett. 2015; 18:1270–1284.View ArticleGoogle Scholar
- Penn AS, Conibear TC, Watson RA, Kraaijeveld AR, Webb JS. Can simpson’s paradox explain co-operation in pseudomonas aeruginosa biofilms?FEMS Immunol Med Microbiol. 2012; 65(2):226–35.View ArticlePubMedGoogle Scholar
- Koella JC. The spatial spread of altruism versus the evolutionary response of egoists. Proc R Soc Lond Ser B Biol Sci. 2000; 267(1456):1979–85.View ArticleGoogle Scholar
- de Boer RJ, Staritsky AD. CASH. 2000. http://bioinformatics.bio.uu.nl/rdb/software.html. Accessed 29 Jan 2016.