Field Notes

entity

Comprehensive School Threat Assessment Guidelines

StatusStableUpdatedStands on4 sourcesSupportsnotes

The Comprehensive School Threat Assessment Guidelines (CSTAG) is the school threat-assessment model developed by Dewey Cornell and colleagues at the University of Virginia in 2001. Originally called the Virginia Student Threat Assessment Guidelines, it was renamed CSTAG in 2018 as its adoption spread beyond Virginia, and it is now the most studied of the school behavioral threat assessment models in United States use. Its distinguishing commitments are a decision tree that resolves most cases quickly, a refusal of zero-tolerance responses to minor threats, and the integration of counseling and mental-health supports into the response to serious ones.1

The decision tree

CSTAG structures an assessment as five steps, designed so that the large majority of reported threats are resolved at the first two.1

  1. Evaluate the threat — interview the student, the target, and witnesses, and record the exact words and circumstances.
  2. Attempt to resolve the threat as transient — if the student retracts, explains, or apologizes convincingly, the case closes as transient, with services added only as needed.
  3. Respond to a substantive threat — protect potential victims, warn the intended victim and parents, seek to resolve the underlying conflict, and discipline only as appropriate.
  4. Conduct a safety evaluation — for very serious substantive threats (threats to kill, rape, or use a weapon), add a mental-health screening and a law-enforcement investigation of planning and preparation, and develop a written safety plan.
  5. Implement and monitor the safety plan.

The model’s load-bearing distinction is between transient threats — jokes, figures of speech, and momentary expressions of anger with no sustained intent to harm — and substantive threats, which always require protective action. In large samples roughly 80 percent of cases are not threats or are transient, about 15 percent are serious substantive, and about 5 percent are very serious substantive. Resolving transient cases without suspension or arrest is the point of the design: it is explicitly nonpunitive, directs discipline toward the underlying conflict, and treats zero-tolerance expulsion as a failure mode rather than a safety measure.2

Adoption and training

Virginia made threat assessment teams mandatory in every public school in 2013 — the first state to do so — under Code of Virginia § 22.1-79.4, which requires teams to include expertise in counseling, instruction, school administration, and law enforcement.3 Florida later adopted CSTAG as its statewide model, and the UVA group summarizes 25 published studies across its history.1 Training runs through Cornell’s School Threat Assessment Consultations organization, built around a manual first published in 2006, renamed with the model in 2018, and issued in a second edition in 2024.2 Training quality is not incidental: Behavioral threat assessment in schools carries the fidelity caveat that the model’s measured benefits concentrate in schools that implement it faithfully.

The evidence base and its limitation

CSTAG is the only school threat-assessment model with causal evaluation. A randomized controlled trial and quasi-experimental studies found that CSTAG schools, relative to control schools, used fewer long-term suspensions, had lower bullying-infraction rates, referred more students to counseling, and showed no racial disparities in who was referred for assessment or disciplined.4 The limitation belongs beside the findings: the Learning Policy Institute’s 2025 review observes that the large majority of CSTAG studies were conducted by UVA researchers where the model originated, so independent replication remains thin even though the internal evidence is consistent.4 The no-disparities result in particular is a property of the model under fidelity, not of threat assessment as such.

  1. Comprehensive School Threat Assessment Guidelines, Youth Violence Project, School of Education and Human Development, University of Virginia. The developers’ own overview and summary of the supporting research, so its account of the evidence is not independent of the model. ↩↩↩

  2. CSTAG manual and decision tree, schoolta.com. The commercial distribution site for the model’s training materials. ↩↩

  3. Code of Virginia § 22.1-79.4, enacted 2013. ↩

  4. Jennifer DePaoli and Stacy Loewe, Keeping Schools Safe? The Research on Behavioral Threat Assessments, Learning Policy Institute, October 2025. An outside synthesis, which is why it can weigh the model’s own research claims. ↩↩

Built on 4 sources (4 external).

Working out connections…