How the flow of activity creates the pathways of thought

In 2004, while riding a bus to the University of Southern California, I spent much of the trip watching rain travel down the window. I was studying psychology and had become intensely interested in the brain, neural plasticity, and the physical processes that could make learning and thought possible.

The droplets did not move down the pane independently of one another. New droplets often entered the wet tracks left by previous droplets. Once a route had become wet, later water was more likely to travel through it. Some parts of the window accumulated well-established streams, while other regions remained relatively dry. A few droplets moved into new territory and formed new pathways, but many were captured by routes that had already been used.

The distribution of water was never exactly the same from one moment to the next. Droplets appeared, moved, merged, separated, and disappeared. Some streams endured while others faded. The total pattern was always changing, yet every new configuration was constrained by the configuration that preceded it.

I sat watching this process for a long time because it seemed to offer a simple analogy for neural activity and plasticity. Electrical and chemical activity in the brain does not move through a neutral network. It moves through a network that has already been shaped by development, experience, and previous activity. When a neural pathway is used, the activity can leave behind changes that influence how activity will propagate in the future.

The essential idea is simple:

Flow leaves a trace, and the trace changes the probability of future flow.

Every droplet was both a traveler and a pathway builder. It was guided by the previous condition of the window, and by moving across the window, it contributed to the condition encountered by later droplets. The water followed history while simultaneously writing new history.

The basic correspondence

The rain window can be used to represent several levels of neural and cognitive organization.

Rain window

Brain and cognition

The window pane

The preexisting neural architecture

Scratches, dust, coatings, and microscopic irregularities

Innate connectivity, developmental organization, and established synaptic weights

New raindrops

Sensory inputs, internally generated activations, or candidate representations

Moving droplets and rivulets

Propagating patterns of neural activity

Wet tracks remaining after a droplet passes

Priming, residual excitability, short-term synaptic facilitation, and recent-memory traces

Frequently reused streams

Strong associations, habits, schemas, and attractor pathways

The entire distribution of water at one moment

The current distributed brain state or working-memory coalition

Streams that continue across moments

State-spanning coactivity

The continually changing overall pattern

Incremental change in state-spanning coactivity

Droplets merging

Neural convergence, binding, coalition formation, and multiassociative search

A large stream capturing nearby drops

Attractor dynamics, attentional capture, and habitual interpretation

A droplet creating a new route

Novel association, exploration, insight, or learning

Evaporation

Decay, forgetting, and loss of priming

Gravity, bus motion, wind, and rain intensity

Goals, motivational state, emotional salience, arousal, and environmental demand

Hydrophobic or obstructed regions

Inhibition, gating, avoidance, and suppressed transitions

The analogy should not be taken to mean that action potentials literally wander through the nervous system in the same way that water wanders across glass. Once initiated, an action potential travels along an existing axon. The analogy is more appropriate at the level of synapses, neural populations, cortical assemblies, and recurrent networks. At that level, previous activation, synaptic strength, inhibition, neuromodulation, and current context influence which downstream populations will become active.

The model is therefore best understood as an analogy for history-dependent flow through a plastic network.

Memory as a change in probability

A wet track stores information in an unusual way. It does not contain a picture or verbal description of the droplet that created it. Nothing on the window explicitly says that water previously traveled through a particular location. The history is instead preserved as a change in probability.

A future droplet is now more likely to follow that route.

This resembles implicit memory. A skill, expectation, association, habit, or unconscious motive does not always exist as an explicit statement inside the brain. It may exist as an increased probability that one neural or mental state will lead to another.

The past can therefore survive in at least two forms. It can survive as content, such as an explicitly remembered event, and it can survive as altered ease of transition, such as a habit, expectation, or automatic association.

The rain window represents the second form particularly well. The wet pathway is both a residue of previous activity and a constraint on future activity. It is simultaneously a memory of the past and a prediction about what will happen next.

The past survives not only as stored information, but as altered ease of transition.

This is closely related to the idea I later developed in Iterative Psychodynamics. Much of the unconscious may be encoded in the transition structure of cognition. It may determine which thoughts, memories, interpretations, emotions, and actions are likely to follow the current state without appearing as an explicit conscious proposition.

Activity and plasticity on different timescales

The window also provides a way to visualize several timescales of neural activity and memory.

The water currently moving down the glass corresponds to immediate neural activity. It is active, visible, and changing rapidly. The wetness remaining after the water has passed corresponds to shorter-lived forms of neural facilitation, priming, and residual activation. More permanent surface properties, such as scratches, residue, texture, or previously established channels, correspond to longer-lasting synaptic and structural organization.

These different layers interact:

• Current activity is guided by both recent wetness and durable surface structure.
• Current activity adds new wetness and may gradually strengthen existing channels.
• Unused wetness evaporates, just as activation and priming decay.
• Frequently traversed routes remain more influential than routes used only once.

The analogy therefore depicts immediate activity, short-term memory, and long-term learning in the same visual field.

In my 2016 article on state-spanning coactivity, I distinguished short-term continuity supported by sustained firing and short-term plasticity, intermediate continuity involving priming and emotion, and longer continuity involving learning and episodic memory. The rain window provides an intuitive image of how these timescales can overlap. Some traces disappear almost immediately, while others continue shaping activity long after the original event has ended.

Continuity through partial preservation

The most important feature of the window was that its pattern was continually changing without becoming completely unrelated to its previous state.

Some streams continued. Some ended. Some new ones appeared. The present distribution inherited part of the previous distribution while adding differences of its own.

This is very close to what I later called state-spanning coactivity, or SSC, and incremental change in state-spanning coactivity, or icSSC.

State-spanning coactivity refers to representations that remain coactive across two or more successive brain states. Incremental change in state-spanning coactivity refers to the gradual turnover of that active population. Some representations remain active, others become inactive, and new representations are added.

The rain-window equivalent is straightforward. A stream continuing through several moments represents SSC. The evolving distribution of old and new streams represents icSSC.

This explains how two states can be different while remaining part of one continuous process. Mental continuity does not require the contents of consciousness to remain fixed. It requires sufficient overlap for each state to inherit context and causal organization from the state before it.

A rivulet can also remain identifiable as the same stream even though the individual water molecules composing it are continually replaced. Its identity lies in the continuity of its organization and trajectory rather than in the permanent retention of the same material.

The conscious stream may work similarly. A concern, goal, thought, or sense of self can remain present even though many of the particular neural events and representational details involved in it are changing.

Identity is carried by causal and organizational inheritance.

Convergence and multiassociative thought

Droplets do not only travel independently. They collide and merge. Several small droplets that are individually unable to move may combine into a larger droplet that begins traveling. Several narrow trickles can converge into a larger stream.

This is a useful analogy for neural summation and for the process I called polyassociativity.

A new thought or representation does not always follow from one preceding representation. Several active representations can jointly converge on a new one. The new representation may be weakly associated with each item considered separately but strongly associated with their combination.

For example, several clues held together in working memory may converge on an answer that none of them would have produced alone. The answer appears after the combined activation crosses a threshold.

The rain-window equivalent is:

Several weak flows converge, acquire sufficient force, and initiate a transition that none could have produced independently.

This also provides an analogy for insight. Separate memories, observations, bodily reactions, and concepts may remain disconnected for a long time. When they finally become active together, their combined influence can create a new interpretation.

The resulting insight may feel sudden even though its components were accumulated gradually.

Established streams as attractors

Once a large rivulet becomes established, it does more than carry the water already inside it. It can capture nearby droplets. Water landing close to the stream may curve toward it because of the local distribution of wetness, surface tension, and slope.

This resembles an attractor in a neural network. An established neural or cognitive pattern can recruit incomplete, noisy, or ambiguous information into a familiar configuration.

This process can be highly adaptive. It allows the brain to recognize objects from partial information, complete familiar patterns, retrieve relevant memories, and respond rapidly without reconstructing every conclusion from the beginning.

It can also become restrictive. An entrenched fear, belief, stereotype, attachment expectation, or emotional pattern may capture ambiguous information and route it into an established interpretation.

A new experience may therefore be processed through an old pathway. The external event changes, but the person’s characteristic mode of interpreting it remains stable.

I think of this as trajectory capture:

A previously established cognitive pathway captures new information and incorporates it into a familiar psychological route.

This idea provides a bridge from neural plasticity to personality. Repeated experiences create recurring channels through which perception, emotion, memory, and action are likely to flow.

In this sense:

Biography becomes drainage.

A personality can be understood partly as a characteristic landscape of cognitive and emotional pathways. Some routes are broad and frequently traversed. Others are narrow, weak, or almost inaccessible. Particular cues repeatedly direct processing toward familiar outcomes.

Habits and schemas as drainage systems

A single wet trail can represent one association. A network of intersecting streams is more like a schema, habit, belief system, or learned behavioral policy.

Once a drainage pattern has been established, new water does not need to rediscover the entire route. The structure of the window guides it efficiently. In the same way, a practiced behavioral or cognitive sequence does not need to be consciously reconstructed every time it is used.

This relates to the process I called implicitization.

A sequence may initially require explicit progression:

A \rightarrow B \rightarrow C

After repetition, A may become capable of producing C without requiring B to become fully conscious:

A \rightarrow C

The intermediate reasoning has been compiled into the structure of the transition. The system preserves the result of previous processing while omitting some of the steps that originally produced it.

On the rain window, the route embodies the history of earlier flow. Water can now follow the channel without recreating the conditions that initially caused the route to form.

This can explain how a consciously learned strategy becomes a habit, how a repeated interpretation becomes an assumption, and how a previously explicit emotional conclusion becomes an automatic reaction.

Competition among pathways

Established streams compete for incoming water. A dominant stream can intercept droplets that might otherwise have created alternative routes. As it captures more water, it becomes increasingly influential.

This creates a positive-feedback process:

1. A pathway receives activity.
2. Its activation makes it more likely to receive further activity.
3. Competing pathways receive less.
4. The dominant pathway becomes increasingly established.

A similar process may occur in attention and thought. Once a concern, interpretation, or emotional theme becomes active, it can recruit associated memories and perceptions. Those additions then strengthen the theme and make further related additions more likely.

This may contribute to rumination. The contents of thought can change while repeatedly returning to the same underlying concern. Different memories, imagined conversations, predictions, and interpretations all flow through one dominant channel.

The rain-window model also provides an analogy for attentional capture. A salient or emotionally charged stream can gather a disproportionate share of incoming activity, leaving alternative pathways relatively dry.

Novel paths, creativity, and exploration

The window was not completely determined by its old streams. Some droplets created new routes.

A novel pathway could appear because a droplet landed far from an existing stream, because several droplets merged, because the rain intensified, because the bus changed direction, or because an obstacle diverted the water.

Comparable events occur in cognition. Novelty, surprise, curiosity, emotional intensity, environmental change, and unusual combinations of representations can redirect neural processing.

A new idea may initially be weak and unstable. If it is revisited repeatedly, however, it can become an established pathway capable of guiding future thought.

This creates a balance between exploitation and exploration.

A system that always follows its most established channels will be efficient but rigid. A system that forms entirely new routes at every moment will be flexible but incoherent. Adaptive cognition requires enough persistence to preserve context and enough variation to discover alternatives.

Too little continuity can produce fragmentation, distractibility, and failure to maintain a goal. Too much continuity can produce fixation, perseveration, rumination, and resistance to new information.

The ideal system preserves some pathways while remaining capable of rerouting.

Adaptive cognition requires enough residual structure to guide thought and enough flexibility to prevent thought from becoming trapped.

The present contains a record of the recent past

At any moment, the rain window contains several temporal layers:

• droplets that have just arrived,
• streams currently moving,
• wet paths whose water has already passed,
• fading tracks that are gradually evaporating,
• and dry regions that have not recently participated.

The visible present is therefore temporally thick. It is not an isolated instant. It contains traces of several preceding moments.

Consciousness may have a comparable structure. The conscious present contains focal contents, active goals, emotional carryover, recently displaced information, primed representations, and latent material capable of returning to the focus of attention.

The window displays more than what is happening now. It displays a short history of what has recently happened and a probability map of what is likely to happen next.

Every state is therefore simultaneously:

• an outcome of previous states,
• a current configuration,
• and a set of constraints on future states.

This is the central logic of iterative updating.

Organization without a central controller

No droplet understands the entire window. No central droplet directs the others. The larger pattern emerges from local interactions among gravity, adhesion, surface tension, wetness, momentum, incoming rain, and the physical structure of the pane.

The process is distributed, but it is not random. It is organized by the interaction between current activity and accumulated structure.

This provides an analogy for how consciousness could be composed from nonconscious processes. Individual neural operations do not need to be conscious. They can recognize features, retrieve associations, assign value, prepare actions, and generate imagery through local interactions. Their outputs can nevertheless combine into a globally organized and temporally continuous cognitive process.

The rain window exhibits organized influence without unified agency. A wet path can steer later droplets without possessing awareness or intention. In the same way, an unconscious disposition can steer thought without constituting a second conscious person inside the brain.

The organized stream emerges from the interaction of many processes that do not individually understand the whole.

The rain window and Iterative Psychodynamics

The analogy also applies to unconscious motives, emotional patterns, and defensive responses.

A motive does not have to exist as a hidden verbal statement. It can exist as a tendency for mental activity to move in particular directions. It can increase the probability that certain cues will be noticed, certain memories retrieved, certain outcomes anticipated, and certain actions prepared.

In this analogy:

Conscious contents are the moving water, while unconscious dispositions are the drainage pattern.

An attachment expectation may act like a deeply established channel. New interpersonal events are directed through it and interpreted according to its structure. A person who expects rejection may repeatedly notice signs of distance, retrieve memories of abandonment, and prepare defensive withdrawal.

The events are new, but the route through which they are processed is old.

Denial can also be understood as a form of blocked or diverted integration. Information may enter one part of the system and affect emotion or behavior without reaching the pathways that would connect it to the explicit self-model. Some regions of the window receive water, while other routes remain obstructed or hydrophobic.

Rationalization resembles the merging of water from several sources. Once droplets have joined a common stream, it becomes difficult to determine where each part originated. A conscious judgment may integrate perception, emotion, memory, motivation, bodily state, and social expectation without preserving an introspectively available record of every causal contribution.

The person experiences the resulting conclusion, but not the complete path through which it was formed.

The changing pane as a transition landscape

At every moment, different regions of the window have different probabilities of receiving and transmitting water. Some routes strongly attract flow. Others resist it. Some are close to a threshold and need only one additional droplet to begin moving.

The pane can therefore be understood as a dynamic transition landscape.

In a cognitive system, the current state provides the active contents, while the learned organization of the brain determines where those contents are likely to lead. The transition landscape is shaped by synaptic weights, memories, habits, emotional associations, attentional priorities, goals, and inhibitory processes.

The system’s history is embedded in the probability distribution over its possible futures.

A thought is therefore not selected from a neutral field. It is selected from a landscape that previous thoughts and experiences have already reshaped.

This may be one of the deepest lessons of the rain-window model. Neural activity does not merely move across a preexisting structure. It continually modifies the structure that will guide later neural activity.

Limits of the analogy

The brain is considerably more complex than rain traveling down a pane of glass. The window is largely two-dimensional and passive. Gravity supplies a dominant direction. The brain is high-dimensional, massively recurrent, bidirectional, and capable of actively changing its own routing through attention, inhibition, emotion, and goal-directed control.

The brain can also reactivate pathways that are no longer receiving external input. It can simulate events, rehearse actions, retrieve distant memories, and intentionally strengthen or weaken patterns. In a sense, the brain can modify the slope of its own landscape while activity is moving across it.

The analogy therefore should not be treated as a literal electrophysiological description. Its value lies in illustrating a general dynamic:

Current flow is shaped by the traces of previous flow, and current flow leaves traces that shape what comes next.

An early image of the Iterative Updating Model

Looking back, I can see that the bus-window experience contained many of the ideas I later developed more formally:

• neural plasticity as history-dependent flow,
• partial preservation across successive states,
• state inheritance,
• incremental turnover,
• short-term and long-term traces,
• convergence among multiple inputs,
• competition among trajectories,
• attractor dynamics,
• novelty through rerouting,
• organization without a central controller,
• and identity through continuous replacement.

More than a decade later, I described related processes in terms of state-spanning coactivity and incremental change in state-spanning coactivity. I later developed the Iterative Updating Model, State-Spanning Workspace Theory, and Iterative Psychodynamics.

The rain window remains a useful image because it presents all of these processes together. It contains activity and memory, stability and change, continuity and novelty, competition and convergence. It shows how a process can remain recognizable even though its components are continually changing.

The distribution of water was never the same from one moment to the next, but it was never independent of the distribution that came before it.

That may also be true of thought.

Related publications

Reser, J. E. (2016). Incremental change in the set of coactive cortical assemblies enables mental continuity. Physiology & Behavior, 167, 222-237. https://doi.org/10.1016/j.physbeh.2016.09.019

Reser, J. E. (2018, May 29). A neural model of working memory and mental continuity: The Iterative Updating Model. Observed Impulse. https://www.observedimpulse.com/2018/05/a-neural-model-of-working-memory-and.html

Reser, J. E. (2022). A cognitive architecture for machine consciousness and artificial superintelligence: Thought is structured by the iterative updating of working memory. arXiv. https://doi.org/10.48550/arXiv.2203.17255

Reser, J. E. (2026, July 10). State-spanning workspace theory of consciousness: Iterative updating, diachronic unity, and the progressive evolution of experience. Iterated Insights. https://iteratedinsights.com/2026/07/10/state-spanning-workspace-theory-of-consciousness/

Reser, J. E. (2026, August 6). Iterative psychodynamics: How distributed unconscious processes compose and shape the conscious stream. Iterated Insights. https://iteratedinsights.com/2026/08/06/iterative-psychodynamics-how-distributed-unconscious-processes-compose-and-shape-the-conscious-stream/