Evolution of diverse (and advanced) cognitive abilities through adaptive fine-tuning of learning and chunking mechanisms

Abstract

The evolution of cognition is frequently discussed as the evolution of cognitive abilities or the evolution of some neuronal structures in the brain. However, since such traits or abilities are often highly complex, understanding their evolution requires explaining how they could have gradually evolved through selection acting on heritable variations in simpler cognitive mechanisms. With this in mind, making use of a previously proposed theory, here, we show how the evolution of cognitive abilities can be captured by the fine-tuning of basic learning mechanisms and, in particular, chunking mechanisms. We use the term chunking broadly for all types of non-elemental learning, claiming that the process by which elements are combined into chunks and associated with other chunks, or elements, is critical for what the brain can do, and that it must be fine-tuned to ecological conditions. We discuss the relevance of this approach to studies in animal cognition, using examples from animal foraging and decision-making, problem-solving and cognitive flexibility. Finally, we explain how even the apparent human-animal gap in sequence learning ability can be explained in terms of different fine-tunings of a similar chunking process.This article is part of the Theo Murphy meeting issue 'Selection shapes diverse animal minds'.

Keywords: animal cognition; cognitive evolution; configural learning; evolution of learning; non-elemental learning; sequence learning.

Figure 1.

Schematic description of how the similar edge structure of two nodes in the network can support generalization. (a) The incoming and outgoing edges of nodes A (red arrows) and Z (blue arrows) have the same structure (both have incoming edges from X and Y, and outgoing edges to B, C and D). (b) As a result of this similar edge profile, using the network for planning a sequence of actions (by some mechanism that tracks the edges and their relative weight), makes it possible to navigate from Y to B and then to ‘food’, through either A (pink path) or Z (light blue path). Similarly, it is also possible to navigate from X to C or to D through either A or Z (not highlighted in the figure). (c) Although a limited set of sequences experienced by the animal (each blue line represents a particular experienced sequence) is sufficient for creating the network described in (a) and (b), the network can then be allowed to produce novel sequences never experienced before (and are therefore not shown in the figure but can be easily visualized): some of them can lead to ‘food’ and may therefore be viewed as functional (XZB-food and YAB-food) and some are novel but do not lead to food (XZC, XAD, YAC and YZD). (d) Rapid chunking may cause A and Z to become embedded within specific chunks such as XA, AD, ZB and ZD, resulting in a network that impairs generalization: the only way to reach food is by following the previously experienced sequences XAB-food and YZB-food, and generating any novel sequences, even non-functional ones, is impossible.

Figure 2.

Schematic description of the process that creates chunks that are either sequence insensitive (cannot represent the sequential order of A and B) or sequence sensitive (that represent this sequential order). A and B are nodes representing the elements A and B. Solid arrows are edges leading from A and B to n, n′ or n″, which are the nodes that will form the chunks {A, B}, AB and BA, respectively (see text for full description). Horizontal dashed arrows indicate whether the sequence is from A to B or from B to A, and the length of the dashed arrow indicates the time difference between the initiations of each event. The red and blue curves represent the activation profiles of the nodes A and B, respectively, while the black curves represent the sum of their activation when these activations overlap in time, and may thus create an activation peak. The activation profile of each element is assumed to increase during its occurrence and decrease (decay) as soon as it ends. (a) The formation of a sequence-insensitive chunk {A, B} is expected when the edges leading from A and B are symmetric, so the time needed for the activation signals to reach n is the same. Repeated peak activations will then create the chunk {A, B}, either when A and B occur simultaneously or nearly simultaneously (left) or one after the other if the decay is slow (right), but not when the activation of the first decays before the activation of the other (centre; no overlapping activation). (b) Reversing the sequential order from AB to BA does not change the peak activation curves (compare with (a)), meaning that the chunk n is sequence insensitive. (c) When the edges leading from A and B are asymmetric, causing the signal from A to arrive later to n′ and the signal from B to arrive later to n″, sequential activations of A followed by B versus B followed by A can result in different peak activations of n′ and n", respectively, which after repeated occurrences create the sequence-sensitive chunks AB (formerly n″) and BA (formerly n″). (d) When the asymmetry in signal-arrival time is less pronounced than in (c), and the decay is slower, the sequence AB results in peak activations in all three nodes, though this peak is higher and broader in n′ (see text for further explanations and implications).