Transitions in cognitive evolution

Abstract

The evolutionary history of animal cognition appears to involve a few major transitions: major changes that opened up new phylogenetic possibilities for cognition. Here, we review and contrast current transitional accounts of cognitive evolution. We discuss how an important feature of an evolutionary transition should be that it changes what is evolvable, so that the possible phenotypic spaces before and after a transition are different. We develop an account of cognitive evolution that focuses on how selection might act on the computational architecture of nervous systems. Selection for operational efficiency or robustness can drive changes in computational architecture that then make new types of cognition evolvable. We propose five major transitions in the evolution of animal nervous systems. Each of these gave rise to a different type of computational architecture that changed the evolvability of a lineage and allowed the evolution of new cognitive capacities. Transitional accounts have value in that they allow a big-picture perspective of macroevolution by focusing on changes that have had major consequences. For cognitive evolution, however, we argue it is most useful to focus on evolutionary changes to the nervous system that changed what is evolvable, rather than to focus on specific cognitive capacities.

Keywords: comparative cognition; major transitions; neural networks; unlimited associative learning.

Figure 1.

Five types of computational architecture. The hydra Hydra vulgaris (a) has a decentralized computational architecture. Its nervous system is a diffuse neural net within which control flow is best described as local and distributed sensorimotor transformation. An example of a centralized nervous system is the flatworm Dugesia japonica (b). The nervous system has bilaterally symmetrical cephalic ganglia which receive input from the major sense organs and coordinate the activity of the body via parallel longitudinal nerve cords. The control flow is dominated by a centralized feed-forward sensorimotor transformation, with the output from the brain delivering commands to coordinate the body (red arrow). (c) A simplified frontal section of an insect brain. Within the insect brain are sensory lobes (optic lobe ol and antennal lobe al), integration centres (mushroom bodies mb and central complex cx) and premotor centres (pm) including the dorsomedial protocerebrum and lateral accessory lobes. The control flow (adapted from [35]) is dominated by bidirectional recurrent loops that connect the sensory and premotor systems (red), hence this is an example of a recurrent computational architecture. (d) Simplified sagittal section of an avian brain. The control flow involves recurrent connections linking the pallium (pa), thalamus (th), tectum (tc), nuclei of the basal ganglia (bg), pontine nuclei (p) and cerebellum (cb). Each of these regions contains recurrent systems (red). In this laminated computational architecture, the control flow of any specific task can be optimized through learning to improve efficiency. (e) Simplified sagittal section of a human brain. In addition to the control flow in (d), there are virtual control representations generated by computation in structures of the cerebral hemispheres (ch) (particularly the frontal and orbitofrontal lobes) that can modify the control flow to improve both the execution of a task and the efficiency of control flow for a specific task. This ability of the human computational architecture to modify its own architecture and control flow is reflection.