Where Thoughts Come From: Environment, Brain States, and the Question of Free Will
One of the most interesting facts about the mind is that I can observe a thought after it appears, but I usually cannot name the next thought before it arrives. Thoughts feel personal and intimate, yet they often arise with a degree of spontaneity that is difficult to explain. That simple observation sits at the center of a much larger question: where do thoughts come from?
My own view is that thoughts are not generated out of nowhere. I suspect they are produced by a lawful interaction between brain state, body state, memory, social context, and environment. In that sense, what feels spontaneous may not be uncaused. It may simply be the product of variables that are too numerous and too complex for us to track in real time. That view leans toward a deterministic picture of mental life, even if I remain open to being proven wrong. At present, science does not prove that all thought is determined in the philosophical sense, but it does strongly support the idea that thoughts are shaped by prior brain activity, personal goals, emotionally significant concerns, and environmental context.
A good place to begin is the science of spontaneous thought. People spend a large portion of waking life in daydreams, internal narration, memory replay, future simulation, and stimulus-independent thought. A 2024 review describes spontaneous thought as thought that is relatively free from deliberate constraints and notes that mind wandering may account for roughly 25 to 47 percent of waking thought. That matters because it means much of human mental life is not tightly scripted by immediate tasks. The mind appears to generate content continuously, even when nothing external is demanding it.
But spontaneous does not mean random in the strong sense. The same review argues that these thoughts are often connected to memory, planning, and adaptation. A 2024 PNAS study on brain decoding found that spontaneous thought is often tied to self-relevance, emotional tone, personal narratives, current concerns, past experiences, and future plans. In other words, the thoughts that seem to “just appear” are often organized around what matters to the person. They may feel unpredictable from the inside, but they are not content-free.
This is where my own hypothesis becomes interesting. If thoughts are shaped by self-relevance, emotional salience, memory, goals, body state, and environment, then in principle a sufficiently advanced model might become able to predict them with increasing accuracy. To do that, it would need much more than one brain scan. It would likely need a vast, changing map of variables: genetics, developmental history, trauma exposure, social status, friendships, current environment, finances, sleep, diet, hormonal state, substance exposure, physical activity, cultural inputs, and moment-to-moment sensory context. The challenge is not merely computational power. It is identifying which variables matter most, how they interact, and on what timescale.
There is already early evidence that the brain contains information about thought content that can be modeled. In the 2024 PNAS study, researchers predicted dimensions of spontaneous thought, especially self-relevance and valence, from fMRI data using personalized narratives. The authors explicitly frame this as a step toward decoding internal thought and context from brain activity. That does not mean scientists can read all thoughts on demand, but it does show that internal mental content leaves measurable signatures in the brain.
Neurotechnology is also moving beyond coarse categories. In severe paralysis, experimental brain-computer interfaces have already translated attempted speech into intelligible language in near real time. A 2024 NEJM report described a rapidly calibrating speech neuroprosthesis that decoded cortical activity associated with attempted speech. This is not full “mind reading,” and it works in highly specialized settings, but it demonstrates an important principle: under constrained conditions, neural activity can be translated into meaningful language output.
So could AI eventually predict a person’s next thought? My answer is that partial prediction looks scientifically plausible; complete prediction remains far beyond current evidence. The science already suggests that thought is not independent of prior causes. Spontaneous thought is linked to default mode, attention, and control networks, and recent reviews increasingly describe the brain as a predictive organ that continuously integrates prior information and current input. However, knowing that thoughts are shaped does not yet mean we can forecast them with high precision in open-ended real life. Human mental life is high-dimensional, context-sensitive, recursive, and self-modifying. Prediction may improve, but the problem is enormously harder than predicting a simple motor output.
This brings us to free will. Many people assume that if thoughts arise from prior causes, free will disappears. Neuroscience has sometimes been used to support that claim, especially through the famous literature on the readiness potential, where brain activity appears to precede conscious awareness of deciding to move. But that interpretation is no longer as straightforward as it was once presented. Aaron Schurger and colleagues proposed that the readiness potential may reflect the averaging of spontaneous neural fluctuations reaching a threshold rather than a hidden unconscious decision fully formed in advance. Schurger has since argued that the classic free-will interpretation of this evidence is on much shakier ground than many people realize. In other words, neuroscience has not settled the philosophical debate.
That nuance matters to me. My own instinct is still that thoughts are generated by lawful causes and that a sufficiently complete model of the person and the environment would predict them better than we can today. But I cannot honestly say that current science proves strict determinism, nor can I say it disproves meaningful agency. What science does show is that thought is deeply conditioned. It emerges from a nervous system that is already active, already predictive, already carrying memories, habits, motives, and bodily signals before a given thought enters awareness.
For psychiatry and psychology, this question is not merely philosophical. If thoughts are shaped by identifiable variables, then mental suffering may also become more predictable. That is one reason I find this topic so important. A future system that integrates brain data, physiology, environment, and life history may not only approximate thought generation more closely. It may also help identify when a person is drifting toward despair, obsession, impulsivity, mania, paranoia, or collapse. The same logic that applies to thought prediction could one day improve prevention in mental health, though this would raise major ethical questions about privacy, autonomy, and consent.
My current position is simple. I do not think thoughts come from nowhere. I think they emerge from a structured system. The fact that I cannot predict my own next thought does not convince me that thought is uncaused. It may only show that I do not have conscious access to the full machinery producing it. Whether that ultimately eliminates free will is still open. But the scientific trend is clear enough to take seriously: thoughts are not isolated events. They are outputs of a living system embedded in a world.
The high-yield takeaway
The best current evidence suggests that thoughts are shaped by prior brain activity, memory, goals, emotional salience, and environmental context. Spontaneous thought is real, but it is not obviously uncaused. AI may eventually predict some aspects of thought with growing accuracy, especially under constrained conditions, but full next-thought prediction in daily life remains speculative. The deeper philosophical question, whether this implies the absence of free will, remains unresolved by current neuroscience.
References Ho N, et al. Why do we think? The dynamics of spontaneous thought reveal its functions. PNAS, 2024. Kim HJ, et al. Brain decoding of spontaneous thought: Predictive modeling of self-relevance and valence using personal narratives. PNAS, 2024. Smallwood J, et al. Spontaneous Thought as an Unconstrained Memory Process. Trends in Cognitive Sciences, 2019. Menon V. 20 years of the default mode network: A review and synthesis. Neuron, 2023. Dimakou A, et al. The predictive nature of spontaneous brain activity across species. Neuron, 2025. Willett FR, et al. An Accurate and Rapidly Calibrating Speech Neuroprosthesis. NEJM, 2024. Schurger A, Sitt JD, Dehaene S. An accumulator model for spontaneous neural activity prior to self-initiated movement. PNAS, 2012. Schurger A. What Is the Readiness Potential? Trends in Cognitive Sciences, 2021. Schurger A. The Neuroscience of Free Will. 2024 lecture summary.
Dr. Dawood Jehangir Togoo