Causation anxiety

Natural selection by itself is not evolution.  It is only a mechanism that, according to Darwin, can lead to evolution.  As Peter and Rosemary Grant put it, natural selection takes place within a generation, but evolution takes place across generations. (Weiner, 1994:79)

*

Over the last few years there has been some debate, mostly between behavioural biologists, about the concepts of causation used in the field. Typically these discussions have focused upon Mayr's (1961) distinction between ultimate and proximate causation, and also Tinbergen's (1963) extension of Lorenz's questions.  According to Tinbergen there are four core questions to resolve when explaining a behaviour, that of ontogeny, phylogeny, function and mechanism.  Some scholars discuss the relationship between Mayr and Tinbergen.

The emergence of niche construction theory and work in evolutionary developmental biology has formed much of the epicentre of recent discussion. For example, Kevin Laland and colleagues have claimed that Mayr sidelined developmental processes and neglected their role in evolution.  Much reference is made to West-Eberhard's (2003) book on evolution, development and their interplay.  Laland's focus upon niche-construction theory has become an arena in which many of these claims are reinforced.

I have been a part of this debate, defending a straightforward Mayrian position. It is not my intention to rehearse all of my previous arguments here, but for the interested reader I suggest looking at Scott-Phillips, Dickins & West (2011) and Dickins & Rahman (2012).  With colleagues I have also engaged in debate with Laland, in Evolution.  There are other papers to be found on my website.

Most recently Laland has published a paper in Behavioural Processes.  Here is the abstract:

Much of what is discussed in this paper has been prefigured in earlier work, which can be found at the Laland Lab site.  My views on all of this are incorporated in the previously referenced papers.  What I want to focus on are claims, in this new paper, about the role of constraint:

Laland introduces the notion of developmental bias.  By developmental bias what is meant (in this paper) is a reduction in the degrees of freedom in phenotypic space. In essence one decision about a phenotype limits the range of options for the next decision-making stage, in development as in evolution. Laland strongly hints at some level of path-dependency here, but falls just short of a full, teleological argument (somewhat oddly, Simon Conway-Morris is not cited in this paper at all.).  Here are two examples:

Artificial selection experiments on Bycyclus butterfly eyespots,combined with evo-devo mechanistic analyses, provide further compelling evidence for developmental bias (Beldade et al., 2002; Allen et al., 2008). Such experiments have established that it is easier to select for both anterior and posterior eyespots to get bigger, or for both to get smaller, than to select for opposite changes in the two pairs of eyespots, leading to a bias for strong positive correlation in eyespot size. Moreover, selection on eyespot color was unable to make one pair of eyespots more gold and the other more black, leading to a bias toward covariation in eyespot color. Similarly, Arthur (2011) points out that there are more than 50 independent origins of legless tetrapods, but only a single (or at most two) incidences of the origin of shelled tetrapods, arguing that the relative frequency of these two types of tetrapod body plan lies not in the relative frequencies with which the selective advantages of a shell and the loss of legs occur, but rather in a bias in the developmental system, such that some kinds of variation are common, others exceedingly rare. (Laland, 2014:3)

Without teleology this becomes an issue of evolvability and possibly of conservation.  Evolution by natural selection is often described as a conservative process, adding to past design rather than starting from scratch.  In many ways this is simply a definitional statement, this is what biological evolution does (discounting extinction etc.). But it also implies that resource allocation will impose constraints on viability. Laland describes the conventional view on constraints:

Such phenomena do not cause evolution, since they do not directly change gene frequencies, they merely shut out some hypothetical areas of phenotype space as ‘out of bounds’ to selection, drift, and so forth. Constraints are not, as generally conceived, causes of evolution – rather they are causes of the absence of evolution. Constraints might, for instance, explain why adaptation has not occurred in a given circumstance, or why phenotypes are not globally optimal. Yet selection, drift, mutation and gene flow would remain the only widely recognized evolutionary processes, and developmental and physical constraints would most commonly be viewed as providing limited restrictions on their action. On this view, constraints might explain, say, why pigs cannot fly, since there exist severe size and weight limitations on animals that can get airborne (Haldane, 1985), but the dazzling diversity of life is testimony to the fact that the phenotype space open to recognized evolutionary processes is vast. This conception, which remains the dominant view within evolutionary biology, is that while constraints might impose modest limitations on selection,it is selection and the other evolutionary processes that explain the rich tapestry of life. Constraints would certainly be recognized as a non-trivial part of the causal explanation for a given pattern of taxonomic diversity. Yet, to put it in statistical terms, I suspect that for most evolutionists it is selection and drift that are conceived of as accounting for the vast bulk of variation in organismal form. (Laland, 2014:3)

Note the commitment to evolution as changes in gene frequencies.

Laland continues:

But what if the constraints were sometimes so potent, regular and systematic that they shaped the course of evolution? What if there were severe limitations on those phenotypes that were physiologically plausible? What if there were only certain ways of building bodies? Then, rather than thinking of selection as free to traverse a vast empty chasm of morphospace, it would be channeled along those limited and narrow pathways that were opened up by the processes of development (Gould, 2000). Under such circumstances, a sizeable portion of the variation in organismal form would be explained by developmental ‘constraints’. Indeed, if that were the case the very term ‘constraint’ would be highly misleading. Developmental processes would be the creative element in evolution, demarcating which forms would occur and which would not, and accounting for why organisms possess the characters that they do. Selection, drift and the like would play a more restricted explanatory role than currently recognized. Given its channeling or directing role, some researchers would argue that we should call such events a ‘developmental bias’, rather than a constraint (Arthur, 2004, 2011; Brakefield, 2006, 2011; Müller, 2007). Note, such developmental bias would be a phenomenon that specified the pattern of phenotypic variation and change in the world, not to mention frequently providing a mechanistic account of the origins of that variation, but yet nonetheless did not meet current definitions of an evolutionary process. It would seem that contemporary evolutionary biology does not provide us with adequate conceptual tools specifying how to think about the causal role of phenomena like developmental bias. (Laland, 2014:3; emphasis my own, to point out where he stops just short of a teleological claim.)

Of course, these extensive quotes capture a rhetorical position, an almost poetic claim for a different form of explanation.  What is needed is a true comparison between a developmental bias view of evolution and an orthodox view, a comparison that evaluates the explanatory virtues of both.  This paper does not offer that, but instead ends with a discussion of niche construction theory which amounts to a claim that this perspective incorporates the views above, in essence those views are the mantra of niche construction theorists.

But, before seeking such a comparison we must note that there is a problem of logic.  Under the orthodox view, evolution is changes in gene frequencies. Selection and drift sift genes, variation is introduced by mutation and migration. Laland accepts the basic definition of evolution.  He then describes constraints in a way that suggests various physical properties impact upon the selection of particular alleles.  At this point, these are sufficient conditions for selection to occur.  If prior development has led to only certain future changes being viable (i.e. enabling the organism to survive and reproduce in lay terms) and others not, then if those changes arise that is selection. Whether or not they are both necessary and sufficient is more or less an empirical matter.  Laland has claimed an opposition here, but I can see none.

The next claim is that developmental biases (1) 'specify' the the pattern of phenotypic variation we encounter and (2) mechanistically produce variation. We can now discount (1) as a short hand for a particular form, or contribution to selection.  But what might (2) mean?

Clearly Laland sees the fundamental job of an evolutionary explanation as that of accounting for changes in gene frequencies.  To this end he must be interested in genetic variation and its impact upon phenotypes, and then selection (wheresoever he focuses his empirical gaze).  If there is no connection then he is engaged in another discipline. Presumably he is not seeking to claim that developmental biases in some way generate genetic variation, but he is stating that developmental biases are involved in producing phenotypes. This last claim is trivially true because it is incorporated in any definition of development.  Moreover, no one has claimed that there are no individual differences in development, and therefore variation between developing individuals.  We understand genes to contribute to that in interaction with ecological factors, mitigated by various methods of control of gene expression (most recently epigenetic processes involving methylation and acetylation have been a core research concern).

To echo an argument made in Dickins & Barton (2013), development and evolution are logically separate processes.  For Laland to make his constraint, or developmental bias claims this must be so.  Equally genetic variation, and variation perhaps best captured by the term developmental perturbations (from the norm), are distinct.  Surely Laland must recognize that mutation could drive different developmental outcomes, which in turn could change constraint space? Or is it this kind of traditional, mechanical reductionism that he seeks to avoid? That developmental trajectories are under selection is a fairly standard view, an idea formally and empirically captured by life history theory and it would be interesting to read an account of this body of work from the perspective of Laland and colleagues.

Elsewhere in this paper Laland takes issue with the notion of control, as used in statements that claim that x is under genetic control. He implies that I have claimed cultural transmission to be under such control. In Dickins & Rahman (2012) we discuss the genetic control or constraints that act on epigenetic mechanisms, something that we followed up in 2013, with a discussion of epigenetic adaptations. But transmission is merely a process, the mechanisms that deliver that process will be built in some way.  Genes will be involved in building brains that can learn and communicate, and those capacities may well be influenced by developmental biases.  It is far from clear to me what the distinction is between control, constraint and bias, other than possibly a colloquial hint at degree of action.  Fundamentally all of them refer to systems designed to process inputs in particular ways, and in so doing to generate information. This is something I have discussed before (Dickins & Dickins, 2008) and I leave it to readers of this blog to mull over, should they wish.