From Perception to Appraisal: Hierarchical Brain Responses to Natural and Built Features in Urban Environments

This study demonstrates that urban environments with higher proportions of greenery elicit more positive subjective appraisals and distinct early neural responses (enhanced P1 amplitudes), while later brain components (P3, LPP) specifically predict higher-level aesthetic evaluations, suggesting a hierarchical processing mechanism where natural elements influence well-being through both immediate perceptual and subsequent evaluative pathways.

The Gist

Imagine your brain is a highly sophisticated security camera system inside a bustling city. This study asks a simple question: What happens inside that camera when it looks at a leafy park versus a concrete parking lot?

The researchers, a team of "Neuro-Urbanists" (scientists who study how city design affects our brains), wanted to understand why green spaces make us feel good and why some city views make us feel stressed. They didn't just ask people how they felt; they hooked 63 people up to EEG machines (which read brain waves) to see exactly when and how their brains reacted to different city scenes.

Here is the breakdown of their findings, translated into everyday language:

1. The "Green" Boost (The Good News)

The Finding: When people looked at photos with more trees and grass compared to buildings, they felt calmer, happier, safer, and more in control.
The Brain Reaction: The moment the brain saw the greenery (about 0.1 seconds after the image appeared), it gave a little "spark" of activity called the P1 wave.
The Analogy: Think of your brain's P1 wave as a greeter at a hotel. When a guest (the greenery) walks in, the greeter immediately lights up and says, "Welcome! We love having you here!" It's an instant, automatic signal that nature is a "safe" and "good" thing, even before you've had time to think about it.

2. The "Sky" Surprise (The Bad News)

The Finding: You might think looking at a big, open blue sky would be relaxing. But in this study, too much sky actually made people feel more stressed and less safe.
The Brain Reaction: Scenes with lots of sky and deep distances made people feel exposed.
The Analogy: Imagine standing in the middle of a giant, empty football stadium with no walls. You can see everything, but you have nowhere to hide. This is the Prospect-Refuge Theory in action. Our ancestors evolved to like environments where they could see danger coming (prospect) but also had a tree or a cave to hide behind (refuge). A wide-open city street with a huge sky feels like being out in the open with no shelter, triggering a subtle "vulnerability alarm."

3. The "Straight Lines" vs. "Curvy Lines"

The Finding: The brain reacts differently to the shapes in a photo.

• Curvy, messy edges (like tree branches or bushes) made people feel happier.
• Straight, rigid edges (like building corners and roads) triggered a different brain signal called the N1 wave.
  The Analogy: If the P1 wave is the "Greeter," the N1 wave is the Security Scanner. When the brain sees straight, rigid lines (buildings), the scanner goes, "Okay, I see a man-made structure. I need to analyze this carefully to figure out what it is." It's a more analytical, slightly more effortful process than the instant "love" the brain has for nature.

4. The "Judgment" Phase (Later Brain Activity)

The Finding: The early brain sparks (P1 and N1) happened so fast they didn't actually predict how much a person liked the scene. However, later brain waves (called P3 and LPP) were the real judges.

• If a scene was beautiful and fascinating, the brain held onto the image longer and processed it with more "sustained attention" (LPP).
• If a scene was ugly or boring, the brain had to work harder to figure it out (P3), almost like it was struggling to find a reason to like it.
  The Analogy: Think of the early brain waves as your first impression when you meet someone. The later waves are the deep conversation that follows. If you have a great conversation (high LPP), you rate the person highly. If you have to struggle to keep the conversation going (high P3), you probably won't rate them as highly.

5. Why This Matters for City Design

The study suggests that our brains are wired to love nature instantly, but they are also wired to feel uneasy when we feel too exposed.

• Don't just build tall buildings: If you build a massive skyscraper with a huge open plaza and a giant sky view, people might feel exposed and stressed, even if it looks "modern."
• Add the "Refuge": To make a city feel safe and happy, you need a mix. You need the "prospect" (open views) but also the "refuge" (trees, bushes, nooks, and crannies) to make people feel protected.

The Bottom Line

Our brains are like smart filters. They instantly recognize nature as "good" and "safe" (the P1 spark). But when we look at a city, our brains also check: "Do I have a place to hide?" If the answer is no (too much sky, too much concrete), we feel stressed.

The researchers conclude that to build happier cities, we shouldn't just focus on adding more green; we need to design spaces that balance openness with coziness, giving our brains both a view and a place to rest.

Technical Summary

1. Problem Statement

Urban environments are increasingly recognized as determinants of mental health, with exposure to built environments linked to elevated stress and diminished well-being compared to natural settings. While previous research in Neuro-Urbanism has established that natural scenes are generally preferred and restorative, significant gaps remain:

• Stimulus Limitations: Prior studies often used extreme contrasts (e.g., dense forest vs. concrete jungle) rather than the mixed, continuous nature of real-world urban scenes.
• Feature Isolation: There is a lack of systematic analysis distinguishing between semantic content (e.g., trees vs. buildings) and low-level visual features (LLVFs) (e.g., edge density, color, spatial depth) that drive rapid "gist" extraction.
• Neural Mechanisms: The temporal hierarchy of brain processing—specifically how early perceptual processing (P1/N1) transitions into later evaluative processes (P3/LPP) in response to urban features—remains unclear.
• Sky Visibility: Previous literature suggests sky visibility is restorative, but this has not been rigorously tested in complex urban contexts where openness might signal exposure or lack of refuge.

2. Methodology

Participants:

• Sample: 63 adults (mean age = 29.16 years; 29 male, 33 female, 1 non-binary).
• Exclusions: 9 participants were excluded due to technical issues, non-compliance, or poor EEG data quality (>20% channels rejected).

Stimuli:

• Source: 56 street-level photographs from Berlin, Germany, derived from a previous dataset (Sander et al., 2024).
• Variation: Images varied in proportions of vegetation, buildings, vehicles, persons, and sky visibility.
• Semantic Segmentation: Images were processed using the SegFormer-B5 model (pre-trained on ADE20K) to quantify pixel proportions of specific classes (greenery, built structures, sky, cars, persons). A Greenery/Built Ratio was calculated to avoid multicollinearity.
• Low-Level Visual Features (LLVFs): Eight features were extracted, including Straight Edge Density (SED), Non-Straight Edge Density (NSED), entropy, color metrics (green hue, saturation), fractal dimension, and mean metric depth (using Depth Anything V2).

Experimental Design:

• Task: A within-subjects design where participants viewed 56 stimuli. Each image was presented 9 times, once for each of 9 subjective rating scales.
• Rating Scales:
  • SAM (Self-Assessment Manikin): Arousal, Valence, Dominance.
  • Likert Scales (1-5): Stress, Openness, Safety, Beauty, Hominess, Fascination.
• Procedure: Each trial involved a word-pair prime (1000ms) $\rightarrow$ fixation cross (500ms) $\rightarrow$ stimulus (3000ms) $\rightarrow$ rating scale. Total duration: ~90 minutes.

Neurophysiological Recording:

• EEG: 64-channel active gel electrodes (BrainProducts), sampled at 500 Hz.
• Preprocessing: Performed using the BeMoBIL Pipeline (EEGLAB). Steps included filtering (0.2–125 Hz), downsampling to 250 Hz, ZapLine-Plus for noise removal, bad channel interpolation, and Adaptive Mixture Independent Component Analysis (AMICA) for artifact removal (eye, muscle).
• ERP Analysis: Event-Related Potentials (ERPs) were analyzed for four components:
  • P1 & N1: Occipital electrodes (POz, Oz, O1, O2) at ~105ms and ~152ms.
  • P3 & LPP: Centro-parietal electrodes (CPz, Pz, P3, P4) at ~364ms and ~535ms.

Statistical Analysis:

• Models: Linear Mixed Models (LMMs) using the lme4 package in R.
• Predictors: Semantic ratios (Greenery/Built, Sky, Car, Person) and LLVFs.
• Covariates: Gender, city size of upbringing, extraversion, emotional stability.
• Correction: Benjamini-Hochberg False Discovery Rate (FDR) applied for multiple comparisons.

3. Key Results

A. Subjective Ratings (Psychological Appraisal)

• Greenery: Higher Greenery/Built ratios consistently predicted positive outcomes: lower arousal/stress, higher valence (happiness), higher perceived control (dominance), and higher ratings for beauty, hominess, and openness.
• Sky & Depth: Contrary to some prior assumptions, larger sky areas and higher mean metric depth were associated with negative outcomes: higher arousal/stress, lower valence, lower safety, lower beauty, and lower fascination. This suggests that in urban contexts, excessive openness may induce feelings of exposure or lack of refuge (Prospect-Refuge Theory).
• LLVFs:
  • Non-Straight Edge Density (NSED): Positively correlated with valence, safety, openness, and fascination.
  • Straight Edge Density (SED): Positively correlated with beauty and fascination.
  • Metric Depth: Strongly predicted higher arousal and stress, and lower valence and beauty.

B. Neural Responses (EEG/ERP)

• Early Processing (P1): The P1 amplitude (approx. 105ms) was significantly enhanced by higher Greenery/Built ratios. This indicates preferential early visual processing of natural elements.
• Early Processing (N1): The N1 amplitude (approx. 152ms) was significantly modulated by Straight Edge Density (SED). Higher SED (characteristic of built environments) led to more pronounced (negative) N1 amplitudes.
• Late Processing (P3 & LPP):
  • Neither P3 nor LPP amplitudes were directly modulated by the semantic or LLVF predictors of the images.
  • However, these later components predicted subjective ratings:
    • P3: Higher amplitudes predicted lower ratings of beauty, openness, and fascination (suggesting effortful processing of less pleasant scenes).
    • LPP: Higher amplitudes predicted higher ratings of safety, beauty, and fascination (suggesting sustained attentional engagement with preferred scenes).

C. Perception-Appraisal Disconnect

• Early visual components (P1/N1) were sensitive to image features but did not predict subjective ratings.
• Late components (P3/LPP) were not sensitive to image features directly but did predict subjective ratings.
• This suggests a hierarchical separation: early processing detects features (natural vs. built), while later processing integrates these into an evaluative judgment.

4. Key Contributions

1. Hierarchical Processing Model: The study provides empirical evidence for a temporal dissociation in urban scene processing:
   • ~100ms (P1): Preferential detection of natural elements (greenery).
   • ~150ms (N1): Sensitivity to geometric/structural features (straight edges of buildings).
   • >300ms (P3/LPP): Evaluative processing that correlates with subjective appraisal but is distinct from the initial feature detection.
2. Re-evaluation of "Openness": The findings challenge the notion that sky visibility is inherently restorative in urban settings. Instead, excessive sky and depth in built environments appear to increase stress and reduce perceived safety, likely due to a lack of "refuge."
3. LLVF vs. Semantic Content: The study demonstrates that while semantic categories (greenery) drive early neural responses, low-level features like edge density and depth are critical drivers of subjective stress and aesthetic preference.
4. Ecological Validity: By using real-world street views rather than synthetic or extreme stimuli, the study offers a more realistic model of how humans process complex, mixed urban environments.

5. Significance and Future Directions

• Urban Design: The results support the integration of greenery to improve well-being but caution against designs that create excessive, unrefined openness (e.g., wide, treeless avenues) which may induce stress.
• Neuro-Urbanism: The study establishes a methodological framework for linking specific visual features to neural markers, moving beyond simple "nature vs. city" dichotomies.
• Limitations & Future Work:
  • The use of static 2D images limits the capture of dynamic, 3D spatial experiences.
  • High multicollinearity between features (e.g., sky and depth) makes isolating causal factors difficult.
  • Recommendations: Future research should employ Virtual Reality (VR) and mobile EEG in situ to validate these findings in dynamic, real-world contexts and to better understand the role of eye-tracking in selective attention to specific scene regions.

In conclusion, this paper elucidates the neural trajectory from the rapid, preferential detection of natural elements to the complex, effortful evaluation of urban scenes, providing a nuanced biological basis for evidence-based urban planning.