Skip to content

Stress

Stress in the context of cognitive monitoring refers to the physiological and psychological response to perceived demands that approach or exceed an individual's perceived coping resources. Unlike cognitive workload - which captures objective mental resource utilisation - stress involves subjective appraisal and emotional reactivity.

Stress vs. Workload

These two states are related but distinct:

Feature Cognitive Workload Stress
Definition Mental demand relative to capacity Perceived threat/demand relative to coping ability
Primary mediator Cognitive resource allocation Autonomic nervous system (SNS activation)
Key EEG signature Frontal theta ↑, alpha suppression Frontal asymmetry, beta ↑
Key peripheral signature HRV ↓ (load-related) HRV ↓ (stronger), cortisol ↑, GSR ↑
Temporal scale Task-aligned (minutes) Can accumulate over hours
Subjective instrument NASA-TLX SAM, STAI

Stress and workload interact: high workload triggers acute stress when the operator perceives that they may not cope; sustained stress impairs cognitive capacity and increases apparent workload.

The Valence–Arousal Framework

Stress is commonly modelled in a two-dimensional affective space:

  • Valence - the pleasantness/unpleasantness of the emotional state (positive to negative).
  • Arousal - the intensity or activation level of the state (calm to excited).

Stress occupies the high arousal, negative valence quadrant. This two-dimensional representation allows continuous tracking of emotional state rather than forcing a binary "stressed/not stressed" classification.

The Self-Assessment Manikin (SAM) is the standard instrument for collecting valence and arousal ratings, using pictographic 9-point scales that are accessible across languages and cultures.

Physiological Correlates

EEG

Feature Association with Stress
Frontal alpha asymmetry (F4 − F3 alpha power) Greater right frontal activation (lower right alpha) correlates with negative affect
Beta power (13–30 Hz) Increases under stress and anxiety
Prefrontal theta Increases with emotional processing demands
Amygdala-linked activity Captured via temporal electrodes (T7, T8)

Autonomic Nervous System

Stress produces strong autonomic signatures via sympathetic nervous system (SNS) activation:

Signal Stress Response
Heart rate (HR) Increases
HRV (HF band) Decreases sharply
LF/HF ratio Increases (sympathetic dominance)
Galvanic skin response (GSR/EDA) Increases; skin conductance rises
Pupil diameter Dilates (LC-NE activation)
Respiratory rate Increases

Speech

  • Fundamental frequency (F0) increases and becomes more variable.
  • Vocal tremor amplitude increases.
  • Speech rate increases under acute time-pressure stress.
  • MFCC features shift as vocal tract tension increases.

Chronic vs. Acute Stress

Two distinct temporal profiles are relevant:

Acute stress - a rapid-onset response to a specific stressor (a traffic conflict, radio failure, emergency). The physiological response peaks within seconds and subsides within minutes if the stressor resolves.

Chronic stress - sustained activation of the stress response over hours or days (shift fatigue, ongoing understaffing). Chronic stress depletes cognitive resources, disrupts sleep, and impairs immune function. Its physiological signature overlaps with fatigue and is harder to distinguish from individual baseline variation.

The Brain FM targets acute stress monitoring in the first instance, using short-window predictions. Chronic stress monitoring requires longer temporal context and session-to-session comparison against individual baselines.

Labelling Instruments

Instrument Format What It Measures
SAM (Self-Assessment Manikin) 9-point valence + 9-point arousal Continuous affect state at measurement time
STAI (State-Trait Anxiety Inventory) 20-item questionnaire State anxiety (current) and trait anxiety (disposition)
PANAS (Positive and Negative Affect Schedule) 20-item, two subscales Positive and negative affect (weekly or current)

UNIVERSE collects Affective Sliders and PANAS alongside NASA-TLX, enabling multi-label training for simultaneous workload and affect prediction.

Datasets with Stress Labels

Dataset Stress/Affect Labels Physiological Modalities
UNIVERSE Affective Sliders, PANAS EEG, EDA, PPG, temp, accel
MOCAS SAM (arousal + valence) EEG, webcam
WAUC - (primarily workload) EEG, ECG, BVP, GSR