Dual-Processes on Dual Dimensions: An Investigation Focusing on Peak-Shift in Humans

Abstract

Human associative learning can be conceptualized as the product of a single, top down propositional system, or as a combination of this and a more automatic, bottom up ‘associative’ system. This is known as the single vs dual-process debate. One recent dual-process account (McLaren et al., 2019) posits that human learning is the product of a single system, but in which both associative and propositional processes operate. It posits that associative processes are fundamental to human learning and that they can be combined in such a way to form the basis of propositional thought. This thesis attempts to investigate this debate, and this specific instantiation of a dual-process account by looking at the Peak Shift effect; an associative learning phenomenon in which subjects show the greatest conditioned responding to a stimulus they have not been trained on, and one which is commonly used in support of a dual-process account.

Chapter 1 introduces and discusses the Peak Shift Effect, the dual and single process debate, and integrates these two areas. Chapter 2 introduces a peak shift paradigm in which stimuli are used that vary on two dimensions, with the aim of promoting propositionally driven responding to one dimension, and associatively driven to the other, providing evidence for both sets of processes. Chapter 3 investigates whether this result can be explained by the ideas of McLaren et al., (2019), by introducing an elemental associative network that is able to simulate both patterns of responding seen in Chapter 2. Chapter 4 investigates whether the effects seen in Chapter 2 can be enhanced by the use of perceptually different stimuli, but show that the effect is removed, and that the sets of stimuli used can have a considerable impact on one another. The results are then simulated using the same network and discussed in light of the overarching debate. Chapter 5 modifies the peak shift paradigm, in an attempt to minimize the performance of propositionally driven behavior, but maximize that of associatively driven behavior, but instead achieves the converse result. Finally, Chapter 6 relates the findings of each respective chapter back to the theoretical questions introduced in Chapter 1.