In the 4th century C.E., a seed was planted in Greece that would come to be cultivated by some of the most prominent minds in history. A philosopher by the name of Aristotle would be the first to postulate that memories are developed on connections between ideas or sensations. Such associations between individual elements of the mind, he argued, were governed by the principles of contiguity, frequency, and similarity. Early British associationism was profoundly influenced by the philosophical work of John Locke, David Hartley, and David Hume, as nativism and empiricism contended amongst one another throughout the Renaissance period¹. Under the influence of Locke and Hartley, William James proposed a model of memory that understood mental associations as networks throughout the brain consisting of interrelated components between experiences. James’ emphasis on the role of familiarity in forming memories was not far removed from Aristotle’s associative principle of similarity, claiming:

“The brightest minds are known for their ability to perceive remote similarities through their ‘electric aptitude for analogy.“

James, the first to postulate that such networks formed malleable physical connections in the brain, effectively cast the die of associationism that would come to be epitomized through Hebbian learning.

While James’ belief that the linkage of memories was made possible through the physical connections in the brain, the physiological mechanisms by which this occurred would not be established for decades¹. During the 19th century, Santiago Ramón y Cajal proposed the Neuron Doctrine and the law of specific connectivity, which would prove to be fundamental catalysts in the next century of learning and memory research. Cajal established that synapses between individual neurons had the ability to strengthen or fade according to experience, termed synaptic plasticity. The idea of learning as pertaining to strong neuronal connections was not entirely new — such a concept was first hypothesized in James’ early associative models. Building on Cajal’s work on synaptic plasticity, Donald Hebb proposed that coactive neurons responding to the same stimuli would result in a strengthened synaptic connection¹. The principle that “neurons that fire together, wire together” provided the theoretical groundwork for Hebbian learning, as the repeated strengthening of a synapse in accordance with familiar stimuli resulted in strengthening neuronal patterns. Hebb’s Principle established the mechanisms behind associative models of memory — associations between events, ideas, or stimuli occur as a result of a similar experience that activate patterns in our neural activity.

The behaviorist movement proved pivotal in the experimental application of associative learning principles, with researchers such as Pavlov, Thorndike, and Skinner demonstrating the role of associations in learned behavior responses through stimuli. Ivan Pavlov’s animal studies on learned behavior would come to be known as classical conditioning, establishing the processes that strengthen and weaken associations between conditioned and unconditioned stimuli ³. This interest in the prediction and modification of behavior has had many implications for understanding the means by which associations are established. Hebb would come to be greatly influenced by Pavlov’s early scientific study of associative learning, contributing to the development of the Hebbian principle. While the role of contiguity (an Aristotelian principle of association) in conditioning initially posed a challenge to Hebb, the 1972 error-correction model postulated by Rescorla and Wagner provided a cohesive explanation for how these processes can operate in respect to one another¹.

Lømo and Anderson discovered long-term potentiation (LTP) by inserting electrodes into presynaptic neurons in the hippocampal region in order to stimulate an action potential⁴. They found that a single high-frequency exposure to a stimuli resulted in long-lasting neuronal changes, where any following weak exposures would lead to a similar heightened response. This has been thought to occur through the modification of presynaptic and postsynaptic neurons — for example, by enhancing presynaptic responsiveness to certain neurotransmitters¹. The role of LTP in the strengthening and modification of synapses, resulting in the formation of neural networks, provides evidence for it as a fundamental mechanism of associative learning and memory. Not only does the discovery of LTP provide a complementary neurobiological basis for Aristotle and James’ associationism, but it proposes a neural mechanism for Hebbian learning that works in tandem with synaptic plasticity⁵. The work of Lømo and Anderson further contributed to our understanding of the critical role of the hippocampus in memory. Support for the hippocampal region’s role in the encoding and retrieval of memory has also been identified through various animal experiments. Tsein and colleagues⁶ demonstrated that the “knocking-out” of the hippocampus in mice drastically alters spatial memory. By genetically modifying the hippocampus and blocking LTP, mice are unable to navigate water mazes that are typically learned with ease. Such studies of hippocampal lesions and genetic modification provide evidence for the necessary role of the hippocampus region and LTP in facilitating memory storage, particularly in retrieving spatial associations.

Associationism has, without a doubt, defined our understanding of what comprises the sublime wonder that is the human capacity for knowledge. The Aristotalian seed of associationism planted in 4th century Greece has been nurtured and cultivated by some of the most prominent and defining minds of psychology and philosophy — from Locke’s Of the Association of Ideas to James’ associative network models, Cajal’s neuron doctrine to Hebbian learning. Associationism provides a general schema that has long served as a foundational principle of learning, however, contemporary discourse in cognitive science suggests that the associative framework is an oversimplification. Trettenbein⁵ highlights two Nature publications from Bannerman⁷ and Saucier & Cain⁸ in which the spatial memory of mice was not inhibited by blocking LTP if they were pre-trained in water mazes. Gallistel⁹ calls attention to the error-correction hypothesis of the Rescorla and Wagner model, addressing the conceptual and methodological problems in their trials regarding temporal pairing, a vital organ of associative theory.

Millenia have since passed Aristotle’s early postulations on associationism — thus, it would be quite unreasonable to assume that the associative learning theory ought to perfectly align with contemporary experimental findings. With these shortcomings in mind, I hesitate to present associationism as an idyllic closed-loop that cohesively encompasses the full nature of learning and memory mechanisms, and shuts the door on the unresolved conflicts and disputes of the field. Rather, I resolve to leave the door ajar and conclude with the words of William James:

“…our science is a drop, our ignorance a sea. Whatever else be certain, this at least is certain—that the world of our present natural knowledge is enveloped in a larger world of some sort of whose residual properties we at present can frame no positive idea.”

William James – The will to believe, and other essays (1897)