Wagner's proposed resolution of the robustness-evolvability paradox holds that phenotypes with large neutral networks in many-to-few genotype-phenotype maps are simultaneously robust and evolvable, and Mayer and Hansen (2017) showed by sampling random Boolean genotype-phenotype maps that this positive relationship holds within maps but reverses across maps, resting on strict assumptions. Their analysis left open the general question of when neutral-network (redundancy) structure genuinely determines phenotype-level evolvability and robustness. This problem asks for the missing foundations: how population-genetic processes distribute genetic variation across complex maps containing multiple connected phenotypes of different fitness (beyond the uniform-distribution or single-network assumptions), under what epistatic conditions neutral genetic changes alter the probability of non-neutral changes so that redundancy matters at all, and whether the robustness-evolvability relationship is direction-dependent across morphospace. The question remains open and actively debated in reviews published in 2025-2026.

Background

Evolvability — the capacity of a biological system to produce heritable phenotypic variation on which natural selection can act — and mutational robustness — the ability to maintain the same phenotype despite genetic change — stand in apparent tension: a system insensitive to mutation would seem unable to generate novel variation. Wagner (2008) proposed an influential resolution rooted in the structure of genotype-phenotype (GP) maps, the functions assigning phenotypes to genotypes. In many biological GP maps the mapping is 'many-to-few': vastly more genotypes than phenotypes, so the genotypes encoding a given phenotype form a neutral network, a set of genotypes linked by single mutations that all express the same phenotype. Phenotypes with large neutral networks are robust, because most mutations of their genotypes stay on the network; such networks also extend widely through genotype space and therefore abut more distinct neighboring phenotypes, making those phenotypes more evolvable (in this literature, the diversity of phenotypes reachable by single mutations). In RNA secondary-structure maps this yields positive correlations between the mutational robustness and the evolvability of phenotypes. Draghi et al. (2010) showed with population models that mutational robustness can facilitate adaptation, but that it can also impede it, with the outcome depending on population size, mutation rate, and the structure of the fitness landscape.

Mayer and Hansen (2017) tested the generality of this 'paradox resolved' by sampling large numbers of random Boolean GP maps. They found a preponderance of positive correlations between evolvability and robustness of phenotypes within a map, but negative correlations between average evolvability and average robustness across maps, and argued the two properties tend to be negatively correlated when phenotypes are measured on ordinal or higher scale types (phenotypes that carry at least a rank order, rather than unordered categories). They concluded that Wagner's positive relationship rests on strict and somewhat unrealistic biological assumptions, thereby 'restoring' the paradox in important settings.

Their analysis closes with three unresolved issues that jointly delimit when map structure actually determines phenotype-level properties. First, the underlying models assume either no within-population variation or variation uniformly distributed over a neutral network, whereas a real population at mutation-selection balance — the equilibrium between recurrent mutation and purifying selection — should sit non-uniformly: van Nimwegen et al. (1999) showed that even purely neutral evolution concentrates populations at the most strongly connected regions of a neutral network (the stationary distribution is the principal eigenvector of the network's adjacency matrix), which reduces mutation load, but only under the assumption that all alternative phenotypes confer the same low fitness. How mutation, drift, and selection jointly distribute variation across complex GP maps containing multiple connected phenotypes of different fitness, and how that distribution feeds back on evolvability, is unknown. Second, it is unclear whether the redundancy of many-to-few maps matters at the phenotype level at all: it matters only under epistasis — dependence of a mutation's effect on the genetic background — in which neutral genetic changes alter the probability or effects of non-neutral changes; if the neutral component of a genotype does not affect which phenotypic changes are accessible, redundancy influences neither robustness nor evolvability. Third, the relationship may be direction-dependent: quantitative-genetic analyses suggest that an intermediate level of pleiotropy (single genes affecting multiple traits) maximizes evolvability averaged over directions in morphospace, the multivariate space of phenotypic measurements (Pavličev and Hansen 2011), but robustness has not been examined as a property that varies across morphological directions, in the way evolvability does.

Problem Statement

Under what conditions does the neutral-network (redundancy) structure of a many-to-few genotype-phenotype map genuinely determine the evolvability and robustness of its phenotypes, rather than the observed positive within-map correlations being consequences of unrealistic assumptions? As left open by Mayer and Hansen (2017), the question turns on three axes: (1) how population-genetic processes — mutation, drift, and selection — distribute a population's genetic variation across a complex genotype-phenotype map that contains multiple connected phenotypes of differing fitness, beyond both the no-variation assumption and the uniform distribution over a single neutral network (van Nimwegen et al. (1999) characterized the single-network equilibrium not as uniform but as concentrated at the most strongly connected genotypes, and only under the equal-low-fitness assumption), and how this distribution feeds back on evolvability; (2) the epistatic conditions — those under which neutral genetic changes alter the probability or effects of non-neutral changes — under which redundancy influences phenotypic evolvability and robustness at all, and what occurs when these conditions fail; and (3) whether the robustness-evolvability relationship is direction-dependent in morphospace, i.e., whether robustness, like evolvability, varies across morphological directions rather than being a single scalar property of a phenotype.

References

  1. Mayer C, Hansen TF (2017). Evolvability and robustness: A paradox restored. Journal of Theoretical Biology 430:78-85. doi:10.1016/j.jtbi.2017.07.004, https://doi.org/10.1016/j.jtbi.2017.07.004
  2. Wagner A (2008). Robustness and evolvability: a paradox resolved. Proceedings of the Royal Society B 275:91-100. doi:10.1098/rspb.2007.1137, https://doi.org/10.1098/rspb.2007.1137
  3. van Nimwegen E, Crutchfield JP, Huynen M (1999). Neutral evolution of mutational robustness. Proceedings of the National Academy of Sciences USA 96:9716-9720. doi:10.1073/pnas.96.17.9716, https://doi.org/10.1073/pnas.96.17.9716
  4. Pavličev M, Hansen TF (2011). Genotype-Phenotype Maps Maximizing Evolvability: Modularity Revisited. Evolutionary Biology 38(4):371-389. doi:10.1007/s11692-011-9136-5, https://doi.org/10.1007/s11692-011-9136-5
  5. Draghi JA, Parsons TL, Wagner GP, Plotkin JB (2010). Mutational robustness can facilitate adaptation. Nature 463:353-355. doi:10.1038/nature08694, https://doi.org/10.1038/nature08694
  6. Pélabon C, Agudelo-Cantero GA, Araya Ajoy YG, Bolstad GH, Cheng C, Galis F, Guillaume F, Haaland TR, Hallgrímsson B, Hansen TF, Holstad A, Houle D, Hunt G, Isaksen A, Milocco L, Mubalegh N, Nuño de la Rosa L, Orzack SH, Porto A, Reid JM, Sztepanacz JL, Undheim EAB, Villegas C, Voje KL, Walling C, Wright J (2025). Evolvability: progress and key questions. BioScience 75(12):1042-1057. doi:10.1093/biosci/biaf111, https://doi.org/10.1093/biosci/biaf111
  7. Hulst RI, Govers SK, Michiels J, Verstrepen KJ, van den Berg P (2026). The evolutionary role of mutational robustness: theoretical insights. Trends in Genetics 42(2):150-162. doi:10.1016/j.tig.2025.10.011, https://doi.org/10.1016/j.tig.2025.10.011
  8. Hardy NB (2025). How does robustness affect evolvability? Evolution 79(10):1996-2006. doi:10.1093/evolut/qpaf116, https://doi.org/10.1093/evolut/qpaf116
  9. Smerlak M (2021). Neutral quasispecies evolution and the maximal entropy random walk. Science Advances 7:eabb2376. doi:10.1126/sciadv.abb2376, https://doi.org/10.1126/sciadv.abb2376
  10. Catalán P, Wagner A, Manrubia S, Cuesta JA (2018). Adding levels of complexity enhances robustness and evolvability in a multilevel genotype-phenotype map. Journal of the Royal Society Interface 15:20170516. doi:10.1098/rsif.2017.0516, https://doi.org/10.1098/rsif.2017.0516
  11. Jiang P, Kreitman M, Reinitz J (2023). The effect of mutational robustness on the evolvability of multicellular organisms and eukaryotic cells. Journal of Evolutionary Biology 36:906-924. doi:10.1111/jeb.14180, https://doi.org/10.1111/jeb.14180
  12. Mehra P, Hintze A (2024). The Role of Pleiotropy and Epistasis on Evolvability and Robustness in a Two-Peak Fitness Landscape. Biology 13:1003. doi:10.3390/biology13121003, https://doi.org/10.3390/biology13121003
  13. Rigato E, Fusco G (2020). Effects of Phenotypic Robustness on Adaptive Evolutionary Dynamics. Evolutionary Biology 47(3):233-239. doi:10.1007/s11692-020-09506-w, https://doi.org/10.1007/s11692-020-09506-w
  14. Petit AJR, Guez J, Le Rouzic A (2023). Correlated stabilizing selection shapes the topology of gene regulatory networks. GENETICS 224(2):iyad065. doi:10.1093/genetics/iyad065, https://doi.org/10.1093/genetics/iyad065