Darwin's theory of natural selection entails two main components, namely, that (1) organisms produce offspring with at least some heritable variation and (2)
that organisms generally produce more offspring than their environment is able to sustain. Given those two components, some variants are bound to be fitter than others in the sense that their offspring are more likely to survive and produce offspring. This difference, in turn, will cause the heritable traits of the fitter variants to spread in the population. Given this, Darwin's most ''dangerous idea'' (Dennett, 1995), one can explain an organism's attributes in terms of the selective pressures that promoted their spread and, hence, their current existence. An enormous number of such adaptational explanations have been proposed. Many stress that natural selection optimized features for specific functions; others emphasize that natural selection tends to produce optimal compromises between competing functions and/or costs (Maynard Smith, 1982). Generally speaking, the explanatory power of these adapta-tional explanations derives solely from natural selection's second step, the sorting of offspring. Generation of the variants that are sorted is usually assumed to be random and, hence, irrelevant to explanations of the phenotype. This 'adaptationist paradigm' (Gould and Lewontin, 1979) has dominated evolutionary theory for most of its history.
In the 1970s and 1980s, however, the adaptationist paradigm was challenged by authors who stressed that the variants available to natural selection may not really be random (Gould and Lewontin, 1979; Alberch, 1982; Maynard Smith et al., 1985). Central to those challenges was the idea that, even if mutations are random at the genetic level, those random genetic mutations are channeled, or filtered, through mechanisms of development that favor the emergence of some phenotypes. Some structures may be impossible for embryos to develop; others are likely to emerge (Alberch, 1982). If this is true, then natural selection chooses not among a random selection of phenotypes but from a structured set that is determined, or at least biased, by the mechanisms of development. This idea is important, because it suggests that development constrains the power of natural selection to set the course of evolutionary change. It threatens natural selection's widely assumed omnipotence. Some authors carried this threat so far as to exhort biologists to halt their search for adaptive scenarios and to research, instead, the 'generative' mechanisms of development (Goodwin, 1984). Fortunately, most evolutionary biologists today seek a more balanced rapprochement of embryology and evolutionary biology (Gilbert et al., 1996; Wagner and Laubichler, 2004).
Specifically, evo-devo biologists today tend to accept the concept that natural selection is the most prominent determinant of who thrives and who dies, no matter how constrained development might be. They also tend to stress that development itself is subject to descent with modification - i.e., evolution - which means that even fairly tight constraints can change. Therefore, explanations couched in terms of natural selection are not antithetical to those involving developmental constraints, but complementary (Striedter, 2005). Still, the synthesis of natural selection and developmental constraints remains uncertain in one key respect: what if the mechanisms of development were shaped by natural selection to produce variants that are much fitter than one would expect by chance? Then the distinction between the generative and selective components of natural selection (see above) would blur. The developmental production of variants would no longer be random with respect to a species' ecology. This hypothesis, which was pushed furthest by Riedl (1977), is interesting and potentially profound, but not yet supported by much evidence.
Brains were historically considered to be shaped by natural selection, unencumbered by developmental constraints. In general, the size and structure of both entire brains and individual brain regions were thought to be optimized. Jerison (1973, p. 8 ), made this idea explicit when he wrote that ''the importance of a function in the life of each species will be reflected by the absolute amount of neural tissue of that function in each species.'' How development produced that fine-tuning was never specified. Presumably, the idea was that genetic mutations could vary the size and structure of individual brain regions freely, leading to steady improvements in fitness until an optimum was reached. Little thought was given to the possibility that brains might be constrained in how they could evolve. However, a few authors proposed that trophic dependencies between interconnected brain regions might cause entire circuits or systems to change size in unison rather than piecemeal (Katz and Lasek, 1978). Such 'epigenetic cascades' (Wilczynski, 1984) might channel evolution (Katz et al., 1981), but they would not constrain natural selection, because the cascades help to optimize functional brain systems by matching the size of interconnected neuronal populations. That is, epi-genetic cascades act not against, but in conjunction with, the optimizing power of natural selection; they are not classical constraints, which may explain why they have rarely been discussed (Finlay et al., 1987).
The idea of brains evolving under a restrictive developmental rule was proclaimed forcefully by Finlay and Darlington (1995). Their argument was founded on the observation that the various major brain regions in mammals scale against absolute brain size with different allometric slopes (Sacher, 1970; Gould, 1975; Jerison, 1989). Although this finding was well established at the time, it had not been explained; it was a scaling rule without a cause. Finlay and Darlington's major contribution was to propose that the height of a region's allometric slope was related to the region's date of birth (i.e., the time at which the region's precursor cells cease to divide), with late-born regions tending to become disproportionately large with increasing brain size. Why does this relationship exist? Finlay and Darlington (1995) showed that their late-equals-large rule emerges naturally if neurogenetic schedules (i.e., the schedules of what regions are born when) are stretched as brains increase in size and compressed when they shrink. This insight, in turn, prompted Finlay and Darlington to hypothesize that brain evolution is constrained to stretch or compress neurogenetic schedules and cannot, in general, delay or advance the birth of individual regions. In other words, even if evolution 'wanted' to increase the size of only one brain region, it would be 'forced' to change also the size of many other brain regions. Thus, Finlay and Darlington argued that development constrains brains to evolve concertedly, rather than mosaically.
Finlay and Darlington's developmental constraint hypothesis has been challenged by various authors, who all pointed out that brains do sometimes evolve mosaically (Barton and Harvey, 2000; Clark et al., 2001; de Winter and Oxnard, 2001; Iwaniuk et al., 2004; Safi and Dechmann, 2005). In addition, Barton (2001) has argued that correlations between region size and absolute brain size are due to functional requirements, rather than developmental constraints. Specifically, Barton (2001, p. 281) reported that the sizes of interconnected brain regions in what he called a functional system exhibited ''significantly correlated evolution after taking variation in a range of other structures and overall brain size into account.'' Finlay et al. (2001) countered that such system-specific evolution may indeed occur, particularly for the so-called limbic system (see also Barton et al., 2003), but that this does not negate the existence of developmental constraints. In a review of this debate, I concluded that most of it may be resolved by arguing that instances of mosaic (and/or system-specific) evolution occur against a background of concerted, developmentally constrained evolution (Striedter, 2005; see Mosaic Evolution of Brain Structure in Mammals). Both Finlay and Barton seem open to this kind of rapprochement (Finlay et al., 2001; Barton, 2006).
The debate on mosaic versus concerted evolution highlights how little we know about the evolution of neural development or, for that matter, about the role that natural selection played in shaping brains. The developmental data used to support Finlay et al.'s (2001) hypothesis came from just 15 species and were collected by several different laboratories, using diverse methodologies. Moreover, the data are limited to dates of neurogenesis. We know virtually nothing about species differences (or similarities) in how large brain regions are prior to neurogenesis, how quickly the regions grow, or how much cell death they endure. Data on these other, relatively neglected aspects of brain development might reveal additional constraints, and they might clarify how regions can evolve mosaically even if neurogenetic schedules are conserved.
Similarly lacking are data on natural selection and the brain. Although several analyses have shown that the size of some brain regions (relative to absolute brain size) correlates with aspects of a species' behavior or ecology (e.g., Clark et al., 2001; de Winter and Oxnard, 2001; Iwaniuk et al., 2004), such correlations are only indirect evidence for natural selection. More direct data are difficult to gather, because direct demonstrations of natural selection at work require measurements of heritability and fitness functions. As it is, we know so little about how selection acts on brains that debates on its potency are bound to erupt. Clearly, more studies must be performed before we can reach firm conclusions about which aspects of brain development and evolution are tightly constrained and which are subject to specific selective pressures.
Was this article helpful?