🔬 physics

From Pen to Palette: Mathematical Analysis of van Gogh's Psychological Trajectory

This study employs computational methods, including sentiment analysis of letters and fractal dimension analysis of paintings, to demonstrate that Vincent van Gogh's psychological trajectory exhibited mathematical signatures of critical slowing down and a correlation between visual complexity and death sentiment prior to major crises.

Original authors: Anna Singley, Eli Goldwyn, Mark Pitzer, Jessica Logue

Published 2026-07-28

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Anna Singley, Eli Goldwyn, Mark Pitzer, Jessica Logue

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are watching a weather forecast. Meteorologists don't just look at the clouds right now; they look for subtle signs that a storm is brewing long before the first drop of rain falls. They watch for "critical slowing down," a fancy term for when a system (like the weather, an ecosystem, or even a person's mood) gets so unstable that it takes longer and longer to bounce back to normal after a small bump. Think of it like a swing: if you push a swing gently, it swings back and forth smoothly. But if the swing is broken or the chain is fraying, it starts to wobble, take longer to settle, and swing wildly before it finally snaps or crashes. Scientists have used this idea to predict when lakes might turn toxic or when the climate might shift, but what if we could use it to understand the mind? This is the question a team of mathematicians, psychologists, and philosophers asked when they decided to treat the life of a famous painter not just as a story, but as a complex system of data waiting to be decoded.

The paper, titled "From Pen to Palette: Mathematical Analysis of Van Gogh's Psychological Trajectory," takes a wild and creative approach to understanding Vincent van Gogh. Instead of just reading his letters or looking at his paintings to guess how he felt, the authors used computers to turn his life into numbers. They analyzed over 700 segments of his letters to measure emotions like anger, sadness, and the "desire for death," and they scanned 569 of his paintings to measure their "visual complexity" using a mathematical concept called fractal dimension (which basically counts how messy or intricate the brushstrokes are). By lining up these two streams of data—words and pictures—they looked for those "critical slowing down" signs: moments where his emotions and his art became erratic, volatile, and slow to recover, hinting that a major crisis was coming.

Here is what they found. The study suggests that van Gogh's psychological system did indeed show mathematical warning signs before his most famous crisis: the incident in December 1888 where he cut off part of his own ear. About four to six months before that event, the data showed a coordinated spike in chaos. His letters and his paintings began to swing wildly together, showing a pattern of "critical slowing down." Specifically, the authors found that as his "desire for death" thoughts became more volatile, the complexity of his brushstrokes also increased. In fact, there was a statistically significant link between how much he thought about death and how complex his paintings became, with a correlation value of $r = 0.193$ (where $p < 0.001$ ). This suggests that when his mind was in a turbulent state, his art became denser and more intricate, perhaps as a way of channeling that energy.

However, the paper is careful to point out that this wasn't a simple story of "sad man paints sad pictures." In the months leading up to the ear incident, van Gogh's letters didn't actually show a huge spike in explicit thoughts about dying. Instead, the data suggests his system was becoming "fearless" and impulsive. He was painting at a frantic, high-energy pace (creating very complex, high-fractal art) while his conscious thoughts about death were actually quite low. This implies that the crisis wasn't just about him feeling sad; it was about his entire psychological system losing its ability to stay steady. The "variance" (or how much his mood and art jumped around) was huge, and the system was oscillating unpredictably, like a car with a broken suspension lurching from side to side.

The study also looked at the very end of his life, in 1890. Unlike the ear incident, which was followed by a period where he seemed to stabilize and recover a bit, the final months before his death showed a different pattern. The instability didn't go away; it just stayed high. The "variance" in his desire for death and his social isolation remained elevated, and the system seemed "locked" in a state of high volatility without the ability to bounce back. The authors suggest this mathematical signature—a system that has lost its capacity to recover from shocks—was different from the chaotic oscillations seen before the ear incident.

One of the most interesting things the paper rules out is the idea that these changes were just random noise or that his art and letters were unrelated. The data shows that the "slowing down" happened at the same time in both his writing and his painting, suggesting they were part of the same underlying psychological engine. The paper also notes that while there is a link between his art and his thoughts, they aren't identical; the correlation is real but modest, meaning they are connected but respond to stress in slightly different ways.

Ultimately, this paper doesn't claim to have "solved" the mystery of van Gogh's life or to have found a magic formula to predict suicide in anyone. Instead, it suggests that by treating a person's life as a complex system of data, we can spot mathematical patterns that might have been invisible to the naked eye. It shows that before a major psychological crisis, a person's world—both what they write and what they create—can start to wobble in a specific, measurable way. For a curious teenager, it's a reminder that even the most tragic stories have a hidden structure, and that sometimes, the math behind the madness can tell us something profound about how the human mind works when it's under pressure.

Technical Summary: From Pen to Palette: Mathematical Analysis of Van Gogh's Psychological Trajectory

Problem Statement
This study addresses the challenge of quantifying the psychological trajectory of a historical figure with unprecedented temporal resolution. While qualitative analyses have long examined Vincent van Gogh's correspondence and artwork for evidence of psychological patterns, there is a lack of computational methods capable of identifying and quantifying signals of emotional instability across his final decade (1880–1890). The authors hypothesize that van Gogh's psychological system may exhibit statistical precursors of "tipping points"—specifically, signatures of Critical Slowing Down (CSD)—detectable through both textual and visual expression channels. This approach is grounded in the medical hypothesis that van Gogh may have suffered from acute intermittent porphyria (AIP), a condition characterized by episodic attacks and high variance in symptom severity, which could contribute to psychological instability.

Methodology
The research employs a multi-modal computational framework integrating digital humanities, computational art history, and historical psychology. The methodology consists of three primary components:

Data Sources: The study utilizes two distinct but temporally aligned datasets:
- Textual Data: 719 text segments extracted from over 900 letters sent between 1872 and 1890.
- Visual Data: 569 paintings with documented completion dates from 1881 to 1890.
- Alignment: A 30-day proximity window was used to match painting and letter data, yielding 524 matched data points with a mean temporal separation of 2.1 days.
Sentiment Analysis (Textual Modality):
- Letters were segmented into paragraphs and processed using a supervised sentiment classification pipeline.
- A Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel, trained on a labeled social media dataset, classified text into seven psychological categories: desire for death, acquired capability, social isolation, anger, drug/alcohol use, stress, and social connection.
- Outputs were transformed into continuous sentiment scores (0–1) using Platt scaling to measure the strength of sentiment.
Fractal Dimension Analysis (Visual Modality):
- High-resolution scans of paintings were preprocessed using a Difference of Gaussians (DoG) filter to isolate brushwork edges, followed by Line Integral Convolution (LIC) and Otsu's thresholding to binarize the images.
- The Minkowski–Bouligand box-counting algorithm was applied to calculate the fractal dimension, serving as a metric for visual complexity. Higher fractal dimensions indicate denser, more irregular visual structures.
Critical Slowing Down (CSD) Analysis:
- Rolling statistics were computed using a one-month optimal window to identify temporal resolution for CSD detection.
- The study analyzed higher-order statistical moments, specifically lag-1 autocorrelation and variance, across three temporal phases relative to major biographical events: baseline (4–6 months prior), CSD phase (2–4 months prior), and stabilization phase (0–2 months prior).
- Variance-of-variance (heterogeneity of variance) was tested using Fligner-Killeen tests to assess the stability of volatility itself.

Key Results

Cross-Modal Correlations: A statistically significant positive correlation was found between visual complexity (fractal dimension) and "desire for death" sentiment ( $r = 0.193, p < 0.001$ ). Conversely, "anger" showed a negative correlation with visual complexity ( $r = -0.105, p = 0.017$ ), suggesting cognitive narrowing during angry episodes.
Temporal Trends: Visual complexity showed a significant decreasing trend over time ( $p < 0.001$ ), indicating a progressive simplification of brushwork. However, the variance of "desire for death" fluctuated substantially, with notable increases during crisis periods.
CSD Signatures Before the 1888 Ear Incident: The period leading up to the December 1888 ear-severing incident provided the clearest evidence of CSD. Multiple dimensions (acquired capability, desire for death, visual complexity, and social metrics) exhibited coordinated increases in volatility and autocorrelation 4–6 months prior to the event.
- Autocorrelation: Both "desire for death" sentiment and fractal dimension showed sharp increases in lag-1 autocorrelation starting in mid-1888, reaching $\rho \approx 0.6$ and $\rho \approx 0.4$ respectively by October–November 1888.
- Variance Dynamics: While variance increased approaching the ear incident, the variance-of-variance was highest in the pre-ear period, indicating a system oscillating unpredictably between states (chaotic instability) rather than following a smooth trajectory.
Divergence in Crisis Mechanisms: The study distinguishes between the ear incident and the final suicide attempt. The ear incident was preceded by extreme erratic oscillations (high variance-of-variance) but relatively low conscious death sentiment. In contrast, the period preceding the 1890 suicide showed sustained high volatility without the temporary recoveries seen after the ear incident, suggesting a system locked in a critical state with lost capacity for recovery.

Significance and Claims
The paper claims that mathematical approaches can successfully detect psychological transition signatures in historical figures through multi-modal analysis. Key contributions include:

Extension of CSD Theory: The study extends Critical Slowing Down theory, traditionally used in ecology and climate science, to individual psychological dynamics, demonstrating that psychological tipping points involve system-wide reorganization rather than isolated symptom changes.
Multi-Modal Validation: The findings reveal that psychological systems exhibit coordinated slowing across fundamentally different expression channels (linguistic vs. visual/motor), providing evidence that observed CSD signatures reflect genuine system dynamics rather than artifacts of a single measurement modality.
Objective Biographical Analysis: The research offers a replicable methodology for providing quantitative validation to biographical accounts of psychological decline, bridging the gap between digital humanities and clinical psychology.
Differentiation of Crisis Types: The analysis suggests that different crisis mechanisms (e.g., the ear incident vs. the suicide) may be distinguished by the stability of their variance patterns, with some crises emerging from chaotic instability and others from entrenched, persistent risk.

The authors conclude that while the correlation between modalities is modest, the shared underlying dynamical properties support the development of computational approaches for mental health research and the detection of early warning signals in complex psychological systems.

More like this