Spontaneous Cognition

While walking home, a sound brings back a conversation. The conversation recalls something left unfinished; a few moments later, we are imagining tomorrow. No one planned the whole sequence, yet its steps are not equally likely. Spontaneous cognition concerns thought that unfolds without being fully directed by a current task or a deliberate plan. Its movement is shaped by memory, context, emotion, and what already occupies the mind.


Calling a thought spontaneous does not mean calling it random. A useful distinction is between deliberate constraints, such as maintaining a chosen goal, and automatic constraints, such as a learned association or a persistent concern. An off-task worry can be tightly constrained despite being uninvited. Conversely, we can deliberately begin imagining something and then let the sequence unfold with little further direction. Spontaneity depends on how thought moves, as well as how it begins.


This leads to a distinction between content and process. Content is what a thought is about: a person, an event, a possible future. Process is its organisation in time: how long it stays, what follows it, and how often it returns. Two people might report the same amount of negative thought, while one passes through many different concerns and the other repeatedly revisits a small set. Counting contents alone would miss this difference.


The two levels also influence each other. A thought about failure may recruit memories, predictions, and interpretations that make another failure-related thought more likely. A content has then become a constraint on its successors. In rumination, the individual sentences can keep changing while the stream remains within a narrow neighbourhood of concerns. Continued mental activity need not imply continued exploration.


How can we describe such movement mathematically? In an associative memory network, representations are collective patterns of activity. Their learned relations help shape a landscape of possible states. Retrieval stabilises a pattern; adaptation can gradually weaken that stability, allowing a transition to a successor. Potts networks offer a minimal setting for studying these latching sequences. They let us separate changes in what is stored from changes in how the system moves through it.


The important quantities are then properties of a trajectory. Residence time measures how long a representation remains active. Transition probabilities describe its likely successors. Recurrence tells us whether the system keeps returning, while the visited repertoire describes how much of the stored landscape it actually reaches. A network can store many memories and still visit only a few. Relations between memories and interactions between subsystems can make some routes easier than others.


More movement is not automatically better. We need enough stability to finish a sentence or pursue a goal, and enough flexibility to change direction when the situation changes. Exploration and exploitation describe this tension: searching among possibilities and making use of a promising one. My interest is in how a memory system regulates this balance, and how the same system can support focused, exploratory, or repetitive dynamics.


There is also a measurement problem. Reports of experience tell us something about what a person encounters, but asking for a report can change the experience. Neural recordings reveal activity, but a change in a brain signal does not identify a particular thought by itself. A model's memory state, a reported concept, and a statistically inferred brain state are different objects. Connecting them requires explicit tests, rather than assuming they are interchangeable.


This perspective motivates my work on both computational models and human neuroscience. I ask how representational structure, adaptation, and constraints change the routes available to thought. It also offers questions for studying contemplative practices: can a thought remain available while becoming less able to dictate everything that follows? Such questions connect the dynamics of memory to cognitive flexibility, while leaving the correspondence between a model and lived experience open to investigation.


← Essays

Copyright © 2026 • Ani • Licensed under CC-BY-NC-NA