This study examines whether trained and untrained raters can reliably identify four regulatory states — Connect, Protect, Collapse, and Restore — in natural language samples. Using a mixed-methods design with 120 text samples rated by 8 independent raters, we found substantial inter-rater reliability (κ = 0.74) and strong convergent validity with existing emotional regulation measures.

Three primary findings emerged: (1) Regulatory states are detectable in natural language with high reliability, (2) The Four-Mode Gradient maps onto established constructs while offering greater granularity, and (3) Self-awareness emerged as the primary differentiator between adaptive and maladaptive regulatory trajectories.

Mixed-methods design combining quantitative inter-rater reliability analysis with qualitative thematic coding. Pre-registered on OSF (osf.io/f4x6y). 120 text samples sourced from naturalistic communication contexts, rated independently by 8 raters across clinical and non-clinical backgrounds.

Results suggest the Four-Mode Gradient framework offers a practical, accessible tool for identifying emotional regulation patterns in everyday communication. Applications include therapeutic assessment, educational emotional literacy programs, and AI-assisted emotional pattern detection.