A replicator-based response

The replicator view is unable to respond adequately to the challenge from the plasticity hypothesis without giving up on the replicator commitment. The reason is that the view cannot account for the phenotypic variation produced by environmental

induction in terms of replicators alone. Some biologists have said that taking plasticity seriously only requires a shift in which part of the evolutionary process we emphasize, rather than a shift in our view of inheritance (Paaby and Rockman 2014)_{, but this response is not}

available to those who hold the replicator view.

This defense of the replicator commitment starts with the concept of cryptic genetic variation. Cryptic genetic variation is genetic variation that does not have

phenotypic effects except under unusual environmental conditions (Paaby and Rockman 2014). Developmental systems can produce the same phenotype from a number of different genetic variants by controlling gene activity, much like a canal channels water

from different sources to the same endpoint. Biologists call this phenomenon canalization (Waddington 1942). Since natural selection cannot act on genetic variation unless it is expressed as phenotypic variation, canalization allows genetic mutations to accumulate in a population without being exposed to selection. These cryptic variants are invisible unless an extreme environment disrupts the mechanisms of canalization.

Invoking cryptic genetic variation is one way of explaining what happens when environmental novelties induce novel phenotypic responses. Even without new genetic variation, variation that was previously hidden in a population is exposed when

environmental conditions change. In the case of tetrapod evolution, new habitats merely allowed a subset of genotypic variants in the population of stem tetrapods to express, for the first time, a potential for developing narrower pectoral bones. What looks like a case of new variation in phenotypes without new variation in replicators is actually the revelation of cryptic replicator variants. The replicator view can thus respond to the challenge from the plasticity hypothesis by saying that cryptic variants, not new environmental conditions, are the sources of novel phenotypes.

This response is correct on two counts, but it does not succeed as a defense of the replicator commitment. It is true that canalization is responsible for the homogenous phenotypic expression of genetic variants. It is also true that stressful environmental conditions can disrupt canalization and that as a result, cryptic genotypic variants can express themselves as novel phenotypes. But it would be a mistake to infer from these two facts that replicators are the sole sources of variation in cases of plastic phenotypic novelty.

To illustrate the mistake, I'll start with a thought experiment in which phenotypic variation comes from a non-replicator source. Then I'll argue that the experiment is analogous to certain real-world situations.

Imagine a population of asexual and genetically identical butterflies.11 If there is any phenotypic variation within the population, it cannot be due to variation in replicators. Initially, all the butterflies lay their eggs and feed on the same species of plant. Developing butterflies have a genetically controlled mechanism that causes them to imprint on whatever plant they hatch on, so when they mature, they lay their eggs on that same kind of plant. But a developmental fluke distorts the imprinting mechanism of one lucky butterfly and causes her to lay her eggs on a new species of plant. This new plant is more nutritious than the plant the rest of the butterflies imprint on. When the lucky butterfly's eggs hatch, they eat the new plant's leaves and reap fitness benefits because the new food source increases their size. The lucky butterfly's offspring also imprint on the new plant, which means that they lay their eggs on plants of the same species. Time passes. The lucky butterfly's descendants continue to feed and reproduce on the new plant species, and the resulting changes in body size and fitness are passed from generation to generation.

This thought experiment demonstrates that there is no necessary connection between evolutionarily significant phenotypic variation and variation in replicators. Replicators are responsible for the butterflies' imprinting mechanisms, but they are also invariant (by stipulation). The source of the phenotypic variation in body size and fitness, then, can only be the leaves of the plants the butterflies eat, and these leaves are not replicators. If they were, they would have to be selected in virtue of the contribution they make to butterfly fitness. But they are not. The leaves that foster larger, fitter butterflies do not benefit as a consequence. It is variation in the plants qua butterfly environment, in combination with plasticity in the original lucky butterfly that produces a new, fitter phenotypic variant.

11 This thought experiment comes from Mameli (2004), though he uses it to illustrate a somewhat different point.

Of course, real populations are not genetically invariant. Certain subpopulations of real populations are, however, invariant with respect to the alleles that contribute to particular phenotypes. These subpopulations are analogues of the lucky butterfly and her descendants because they share the relevant alleles with respect to a trait of interest, yet can develop different trait values, depending on the environmental conditions they encounter. Within these subpopulations, there is variation in phenotype without corresponding variation in the relevant replicators.

Such subpopulations are the appropriate unit for discussing the challenge from the plasticity hypothesis to the replicator view. The points about cryptic genetic variation and canalization are true, but they only show that new phenotypic variation is sometimes the result of cryptic genetic variation on the scale of a whole population. There are still subsets of populations that share the relevant replicators yet develop different phenotypes in different environments. As long as this is true, then the challenge to the replicator view still stands.