The target article does an impressive job listing phenotypic traits that cluster in lineages with complex cognition. We particularly welcome the authors’ detailed focus on types of embodiment and the way an increasingly complex body demands increasingly complex cognition. These are factors that are easily overlooked. The authors seek to explain how complex cognition evolved in three lineages (vertebrates, euarthropods, and cephalopod mollusks). They argue that these lineages followed similar evolutionary paths: each underwent four major transitions resulting in an increased level of complexity and culminating in the complex cognition we see today.

The concept of a major transition is closely associated with the work of Maynard Smith and Szathmáry (Reference Maynard Smith and Szathmáry1995); it is notable precisely because it allows one to organise the evolution of lineages into a relatively small number of comparable jumps. Any transitional account needs a story about why the changes postulated are transitions: breaks with the past that open up downstream phenotypic space. For Maynard Smith and Szathmáry, major transitions involved changes in the mechanisms of evolution. Others emphasise the dramatic shift in forms of life that follows a major transitional event (Okasha, Reference Okasha2022).

The target article, however, fails to justify why its identified shifts should be considered major transitions in evolution. Indeed, the impression one gets from (say) Figure three is of the steady, gradual accumulation of useful traits across lineages, culminating in complex cognition. Many of the examples have a gradualist flavour as well: they appeal to the evolution of eyes, for example, a classic gradualist story.

Perhaps, gradualism is sufficient for many aspects of their project – again, it may be enough to show that complex cognition co-evolves with complex bodies – but we think it is a missed opportunity. Transitional thinking, at its best, explains why lineages change in such a dramatic way. The explanatory demands are often subtle – downstream benefits to the phenotypic space of one’s descendants is the striking part of a transition, but cannot be the cause of it (as Trestman (Reference Trestman2013) emphasises in other work). Lacking such a story, one runs the risk of cataloguing mere “clutters” of co-occurring properties, rather than true clusters held together by a common causal origin (Halina, Harrison, & Klein, Reference Halina, Harrison and Klein2022). While many traits may reliably co-occur, the correlation doesn’t itself explain the transitions we observe.

If one did want to defend transitions (and we do), there is an important asymmetry between bodies and brains. A more complex body isn’t very useful without a more complex nervous system to drive it. Indeed, this point is illustrated by Coombs and Trestman throughout their article. For example, they write: “Without sufficient sensory information and brain capacity to guide them, flexible limbs would have been of little functional use for complex, three dimensional motion” (p. 16). How then can we explain transitions in brain capacity without referring to later advantages, such as the use of flexible limbs for movement?

We think one can do this by showing how changes in brain architecture can be favoured on resource-usage grounds alone (Barron, Halina, & Klein, Reference Barron, Halina and Klein2023). Since neural systems are under intense selective pressure to conserve energy (Sterling & Laughlin, Reference Sterling and Laughlin2015), doing the same thing more efficiently can be a powerful driver. A move to recurrent neural processing, for example, can evolve on simple resource grounds (Klein, Reference Klein2022). Recurrence may not have any obvious effects at first – but it opens up a host of possibilities for downstream cognitive power, because recurrent systems are computationally more powerful (Klein & Barron, Reference Klein and Barron2024).

Drawing on the above considerations, we have argued for five major transitions in the evolution of cognition: the appearance of diffuse nervous systems, centralisation, recurrence, lamination, and reflection (Barron et al., Reference Barron, Halina and Klein2023). We are pleased that the first three of these, at least, correspond to Coombs and Trestman’s neural transitions; this strikes us as convergent evidence that we’re all tracking the same phenomenon. But our account has neural evolution driving the bus. Unremarkable selection for efficiency may trigger transitions in nervous systems. But, once established, these transitions fundamentally alter phenotypic space for cognitive evolution by expanding computational possibilities and enhancing evolvability via preserved genetic variation. For example, centralised neural networks need to be in place to control large bodies, as well as coordinate muscled limbs and glandular activity of the body as it gets larger. This account not only explains major trait transitions without appealing to downstream effects but also clarifies the co-occurrence of traits, showing why they form a natural cluster rather than a clutter (Halina et al., Reference Halina, Harrison and Klein2022).

We also think that focusing on structural and functional changes in nervous systems allows for a more sophisticated classification of complex cognition. The target article identifies euarthropods, cephalopods, and vertebrates as capable of complex cognition, but makes no finer distinctions. It lists all as capable of tool use, for example. We think there are important distinctions between the type of tool-use seen in humans, primates, and corvids from that seen in cephalopods and euarthropods. In the former case, tool-use involves learning and operating with complex motor sequences to change an object by stages into something fit for a new purpose (e.g., constructing a stone tool, or a stick for fishing out insect larvae). In the latter case, it involves carrying something nice (e.g., a coconut shell or absorbent material). Complex cognition is itself complex; focus on the explanation of transitions can help bring this out.

Our proposed transitions may not be correct or complete. Arguments about the proper number of transitions have been a feature of the literature since Maynard Smith and Szathmáry’s book appeared (Calcott & Sterelny, Reference Calcott and Sterelny2011), and other carvings of the history of life have been proposed (O’Malley & Powell, Reference O’Malley and Powell2016). It may be that Coombs and Trestman are ultimately offering a different but complementary set of changes. However, we believe the central task of any transitional account is to explain both what changed and why. Only then can we bring order to the bewildering variety of changes characterising the evolution of complex cognition.