Cognitive capacity and control in the evolution of intelligence

Abstract A diversity of intelligences arises from the constraints under which animals evolve. However, characterizing how constraints shape intelligence is challenging because it requires relating the restrictions on cognitive mechanisms to those that affect their evolution. We demonstrate the potentially complex interaction between constraints by considering the case study of working memory. Here, information-processing capability is limited by the storage capacity available to hold representations, and the degree of control over those representations. We present an evolutionary model that is mechanistically detailed enough to capture the interactions between capacity and control. This allows us to make quantitative predictions about the distinct patterns of information processing that might be observed across animals. Further, our model’s cognitive detail allows us to fit recall performance on the retro-cue task, illustrating how model predictions can be tested by comparing humans and rhesus monkeys (Macaca mulatta). We find that capacity and control are synergistic and amplify each other’s effects. However, evolution prioritizes investment in capacity because it is required for control to be effective. The strength of synergy varies due to interactions between these cognitive components depending on task complexity, cue reliability, and the availability of metabolic energy. Consequently, our model predicts diversity in investment in capacity and control across animals, and identifies a small number of regimes into which lineages could evolve. We discuss how the computational structure of tasks exerts selection on cognitive designs. Significance Statement Theories about the cognitive processing required for intelligence have been developed largely independently from analyses of evolutionary pressures. This produces an impediment to understanding the diversity of intelligences across animals. We present a mathematical model that bridges these two theoretical domains, focusing on working memory. Our analysis reveals a rich interaction between the capacity to store information and the ability to control that process. We predict that animal intelligences fall into a few major cognitive regimes: initially capacity is prioritized, then capacity and control increase together due to synergy, and finally capacity returns to the fore as adding further control yields diminishing returns. We demonstrate our model makes predictions that can be compared against empirical data via cross-species experiments. Competing Interest Statement The authors have declared no competing interest. Footnotes C.T. Conceptualization, Evolutionary modeling, Methodology, Statistical modeling, Writing – original draft. E.R. Conceptualization, Statistical modeling, Methodology, Writing – review and editing. A.S., N.V., T.M., T.G. Conceptualization, Funding acquisition, Writing – review and editing. E.M. Conceptualization, Writing – review and editing. No competing interests to declare. https://github.com/evanrussek/Capacity_and_Control_Experiment_Analysis