A greater understanding of the emergence of complex brains can be reached by considering the evolutionary function of cognition. Coombs and Trestman provide a tremendous overview of the distribution of neural traits across animals. However, their focus on collating traits leads them to a proposal about how complex brains emerge based on the animal’s form, but says little about the selection pressures driving brain elaboration. With regard to Tinbergen’s (Reference Tinbergen1963) four questions, their mode of explanation is phylogenetic, emphasizing the evolutionary history of organismal form. We propose that this account suffers from the absence of a functional explanation and suggest that the environmental complexity hypothesis provides a complementary theory about the adaptive purpose of nervous systems (Godfrey-Smith, Reference Godfrey-Smith1998; Levins, Reference Levins1968; Sol, Reference Sol2009; Turner, Morgan, & Griffiths, Reference Turner, Morgan and Griffiths2024) .

We argue that brain complexity is expected to coevolve alongside sophisticated sensors and effectors, but that this must be a worthwhile metabolic investment to be favored. Coombs and Trestman relay that animals with complex brains have traits such as high-resolution eyes and flexible limbs. However, since these metabolically costly additions must pay-for-themselves in terms of fitness, this raises the question of their adaptive value (Sterling & Laughlin, Reference Sterling and Laughlin2015) . Indeed, as we describe below, a toy model illustrates that a greater number of sensors and effectors implies more connections between them (Fig. 1). This is analogous to noting that a house with many switches and lights must have a lot of wire in the walls. Let $s$ be the number of sensory receptors, $a$ the number of motor neuron effectors, and ${h_i}$ the number of interneurons in the connecting hidden layer $i$ of a neural network with $n$ layers. Therefore, the total number of connections is $s{h_i} + \mathop \sum \nolimits_{i=1}^{n - 1} {h_i}{h_{i+1}} + {h_n}a$ . Now conveniently assume ${h_i}$ can be characterized by an average, with layers being uncorrelated. This makes it easy to see that inter-brain connections increase with the typical number of neurons within a layer (width $\bar h$ ) or number of layers (depth $n$ ): $\mathop{E}\left[ {\mathop \sum \nolimits_{i=1}^{n - 1} {h_i}{h_{i+1}}} \right] = \left( {n - 1} \right)\bar h$ . A well-known result in artificial neural network theory is that as connections increase as the network becomes richer in the ways it can map inputs to outputs (namely the Vapnik-Chervonenkis dimension bound; Bartlett & Maass, Reference Bartlett, Maass and Arbib2003) . We turn this idea on its head to argue that animals have more connections in order to detect fine-grained patterns and allow intricate responses. Formally, the brain complexity should be an increasing function of sensory and motor system complexity (e.g. $\partial \bar h\left( s \right)/\partial s$ for all $s$ ). Natural brains have many reciprocal connections both between and within layers making the reality far messier. Nonetheless, our toy model demonstrates that complex brains are only useful alongside complex bodies, so the co-occurrence of traits found by Coombs and Trestman is precisely what we should expect. However, this conclusion only intensifies the issue of explaining why these systems were elaborated by natural selection in the first place.

Figure 1.

An example fully-connected feed-forward neural network with two hidden layers $\left( {s=4,{h_1}=2,{h_2}=3,a=2} \right)$ .

Evolution invests in complex brains to allow animals to cope with a heterogeneous and changing environment. Investment in cognition is hypothesized to occur when the environment presents many states that vary drastically in their outcomes for fitness depending on the animal’s actions, so that the environment is complex (Turner et al., Reference Turner, Morgan and Griffiths2024) . Further, there must be reliable information indicating states, such that the animal can often use cues to produce an appropriate action. For instance, consider the evolution of the special-purpose eyes that allow comb jellies (Copula sivickisi) to hunt bioluminescent plankton (Garm et al., Reference Garm, Bielecki, Petie and Nilsson2016). First, the comb jellies’ environment is complex because plankton exist in patches, and substantial energy is lost if these patches are not found. Second, bioluminescent light reliably indicates the location of plankton. Therefore, conditions favor investing in systems for detecting, processing, and propelling toward information indicating plankton.

Considering adaptive function in terms of environmental complexity and the reliability of information allows us to deepen our understanding of the evolution of complex brains. For instance, Coombs and Trestman note that the Cambrian was marked by the emergence of image-forming eyes and more flexible appendages. The environmental complexity hypothesis suggests further theorizing. Perhaps rising ambient light due to ocean oxygenation during the Cambrian enabled predators and prey to reliably detect each other, while the behavior of other species created a complex environment with high-stakes interactions. Analyzing environmental complexity provides an understanding of why lineages move from simple to complex brains, which is significant for interpreting Coombs and Trestman’s history of morphological forms.

Understanding how evolutionary function and morphological mechanisms interact remains a particular challenge for theory on the evolution of nervous systems. Another way to categorize the mode of explanation employed by Coombs and Trestman is as internalist. That is, their study implies traits must be bundled together to be successful, creating an internal influence on the course of evolution (Sterelny, Reference Sterelny1997) . For instance, nondirectional photoreceptors provide the conditions that favor directional pit eyes. By contrast, the environmental complexity hypothesis is externalist, and it focuses on selection coming from the environment. However, a subtle point is that what is considered “the environment” actually arises from how an organism interacts with the world around it; what states an animal can detect, their available actions, and what constitutes a reward. These factors are in fact determined by the bundle of traits possessed by a lineage and evolve over time. Therefore, theory on nervous systems faces the same issue as evolutionary theory broadly: combining internalist and externalist explanations (Laland et al., Reference Laland, Uller, Feldman, Sterelny, Müller, Moczek, Jablonka and Odling-Smee2015) . This unresolved tension puts particular strain on the study of brains because their very purpose is to respond to the external world.