Violence Risk Assessment Expert Witness
A violence risk assessment expert witness is worth retaining only if the opinion is bounded: risk of what, over what period, under what conditions, and how wrong the estimate could be. Dr. Mendonsa uses structured methods and puts those limits in the report.
Violence risk assessment estimates the nature, likelihood, and manageability of future violence. Done properly it is structured professional judgment — the evaluator rates empirically supported risk factors, formulates how violence would plausibly occur for this person, and states what would reduce it. It does not yield a reliable percentage for one individual.
A violence risk assessment expert witness works in one of three traditions, and the difference shows up on cross-examination. Unstructured clinical judgment is the evaluator’s unaided impression. It performs worst, and the field has moved away from it: Hanson and Morton-Bourgon’s meta-analysis of 118 prediction studies (Psychological Assessment, 2009) found empirically derived actuarial measures more accurate than unstructured professional judgment for sexual, violent, and general recidivism alike, with structured judgment falling in between. Actuarial instruments score fixed items and map the total onto an outcome rate observed in a development sample. Structured professional judgment requires the evaluator to rate a defined list of empirically supported factors and then to formulate, state, and defend a final opinion and a management plan. Dr. Mendonsa works primarily in the third tradition and treats actuarial output as one input to it.
What the numbers can and cannot carry
An actuarial score is a statement about a group. Moving from that group to the person in the courtroom is an inference, and a contested one: Hart, Michie and Cooke (British Journal of Psychiatry, 2007, 190 (suppl. 49), s60–s65) reported that 95% confidence intervals around individual-level estimates were so wide as to render them virtually meaningless, a conclusion other researchers have disputed in print. Base rates compound it — where serious violence is rare in the relevant population, even an accurate instrument produces many false positives. So a responsible report gives a risk level with the reasoning exposed, the comparison group named, and the time frame stated. It does not give a bare percentage for one person.
Where these opinions are used
Sentencing and mitigation; parole suitability and release planning; Sexually Violent Predator proceedings under Welfare and Institutions Code § 6600 and mentally disordered offender proceedings under Penal Code § 2962; civil commitment under the Lanterman-Petris-Short Act, which begins with the 72-hour hold in Welfare and Institutions Code § 5150; threat assessment and workplace violence for employers and public agencies; and duty-to-protect litigation. On that last point, California narrowed Tarasoff v. Regents of the University of California (1976) 17 Cal.3d 425 by statute. Civil Code § 43.92 shields a psychotherapist from monetary liability except where the patient has communicated a serious threat of physical violence against a reasonably identifiable victim, and the duty is discharged by making reasonable efforts to communicate the threat to the victim and to a law enforcement agency. Ewing v. Goldstein (2004) 120 Cal.App.4th 807 held that a threat relayed by a patient’s immediate family member can trigger it.
A defensible report names the referral question and the legal standard it is measured against, lists every source relied on, reports response-style and symptom-validity findings, identifies each instrument and why it was chosen, separates static from dynamic factors, sets out scenarios describing how violence would plausibly occur, and closes with management recommendations tied to those scenarios. Candor about what the method cannot do is not a weakness in a risk report. It is the part that survives cross-examination.
The instruments, and what each one is actually for.
No instrument is selected because it is familiar; each answers a specific question, and each has a stated ceiling.
HCR-20 Version 3
Twenty empirically supported violence risk factors across historical, clinical, and risk-management domains. It is a structured professional judgment tool, not a scoring machine: the product is a reasoned formulation and a management plan, not a total score to be read aloud in court.
HCR-20 V3 · SPJVRAG-R
A twelve-item actuarial measure of violent recidivism. It places a person in a score range with an outcome rate observed in the development sample. That is context, not a prediction about the individual, and the report says so in those words.
VRAG-R · ActuarialLS/CMI
Level of Service/Case Management Inventory. General recidivism risk together with criminogenic need — criminal history, education and employment, family, associates, substance use, antisocial attitudes. Built for supervision and case planning. It answers what to work on, not how dangerous someone is.
LS/CMI · Risk-NeedSTABLE-2007
Changeable risk factors relevant to sexual recidivism: significant social influences, intimacy deficits, sexual self-regulation, general self-regulation, and cooperation with supervision. Designed to be read alongside a static measure rather than on its own, and useful precisely because the factors can move.
STABLE-2007 · DynamicPCL-R
Hare Psychopathy Checklist-Revised, second edition. Twenty items rated from a semi-structured interview plus collateral records. Administered only where psychopathy is genuinely at issue, because the label carries more weight with a jury than the score can support.
PCL-RResponse Style and Symptom Validity
A risk opinion is only as good as the presentation it rests on, and examinees distort in both directions. Structured validity testing runs on every case, for exaggeration and for minimization alike.
MMPI-3 · PAI · SIRS-2 · TOMMHow a risk evaluation is built.
The sequence is designed so that the opinion can be traced back to its sources, item by item, under cross-examination.
Define the question
Risk of what behavior, toward whom, over what period, in what setting, and for what decision. A parole release question and a workplace threat question are not the same assessment, and an opinion written for one does not transfer to the other. If the question cannot be stated in those terms, the case is not yet ready for an evaluator.
Assemble the record
Criminal history and charging documents, institutional discipline and incident records, prior treatment and psychiatric records, supervision and probation notes, prior risk assessments, and collateral interviews where available. Self-report alone cannot support a risk opinion, and an evaluation built on it will not hold.
Interview and validity testing
A structured clinical interview covering violence history, ideation, grievance, weapons access, substance use, and protective factors, paired with formal response-style and symptom-validity measures. Findings are reported whether or not they help the retaining party.
Select and score instruments
Instruments are chosen for the referral question and the population, and the choice is justified in the report. Where an instrument has not been validated on someone like this examinee, that is stated rather than papered over.
Formulate and plan
Static and dynamic factors are separated. One or more scenarios describe how violence would plausibly unfold for this person, and each is paired with the conditions, supervision, or treatment that would make it less likely. This is the part of the report that is useful after the hearing ends.
Report, deposition, testimony
A written opinion with the reasoning exposed, then preparation with counsel, deposition, and testimony. Where an opposing evaluator has over-read an actuarial score, the rebuttal analysis addresses the inference rather than the arithmetic.
The background this opinion rests on.
Risk opinions age badly when the evaluator has never worked inside the systems that have to manage the person afterward.
Risk questions counsel ask before retaining.
Will you give a percentage likelihood that this person commits violence?
No, and you should be wary of an expert who will. An actuarial instrument reports an outcome rate for a group of people who scored in the same range in a development sample. Applying that rate to one individual is an inference the data do not fully support.
Hart, Michie and Cooke (British Journal of Psychiatry, 2007) found the 95% confidence intervals around individual-level actuarial estimates so wide as to render those estimates virtually meaningless. Other researchers have pushed back on that analysis, and the argument is still live. What is not in dispute is that a single-number prediction about one person overstates what the method can deliver. The report gives a risk level, names the comparison group and the time frame, and explains the reasoning that produced it.
Which approach do you use, and why not just the actuarial score?
Structured professional judgment is the primary method, with actuarial results as one input. The HCR-20 Version 3 is the usual backbone for general violence, supplemented by the VRAG-R where an actuarial anchor is useful, the LS/CMI where the question is criminogenic need and supervision planning, the STABLE-2007 where sexual risk is at issue, and the PCL-R only where psychopathy genuinely matters to the decision.
The reason not to stop at an actuarial number is practical. An actuarial score is fixed and historical; it tells a parole board or a judge nothing about what would reduce the risk. Structured professional judgment produces factors that can change, scenarios that can be planned around, and recommendations someone can act on.
Can a risk opinion support or rebut an SVP or MDO petition?
Yes. Sexually Violent Predator proceedings under Welfare and Institutions Code § 6600 turn on whether a diagnosed mental disorder makes the person likely to engage in sexually violent criminal behavior. Mentally disordered offender proceedings under Penal Code § 2962 turn on a severe mental health disorder that was a cause of or aggravating factor in the commitment offense, is not in remission, and leaves the person a substantial danger of physical harm to others.
Both statutes require a link between a specific disorder and a specific risk, which is where these petitions most often fail or succeed. Dr. Mendonsa conducted evaluations of this kind as Chief Psychologist with CDCR and accepts retentions from either side, including rebuttal review of an opposing evaluator’s report.
We have an employee making threats. What can a psychologist do that HR cannot?
Threat assessment is a different exercise from a fitness-for-duty examination or a disciplinary investigation, though a case often needs more than one. The question is whether this person is on a pathway toward targeted violence, what is driving it, and what intervention would move them off it.
That work includes structured review of the communications and behavior actually reported, an evaluation where the employee can be examined, and written recommendations about monitoring, workplace measures, referral, and the conditions under which the assessment should be revisited. Dr. Mendonsa also conducts fitness-for-duty evaluations for law enforcement, fire, and other security-sensitive roles.
Does a Tarasoff duty apply, and what does California actually require?
Tarasoff v. Regents of the University of California (1976) 17 Cal.3d 425 established that a therapist who determines, or under professional standards should determine, that a patient presents a serious danger of violence to another incurs an obligation to use reasonable care to protect the intended victim.
California then narrowed it by statute. Civil Code § 43.92 provides that there is no monetary liability for failing to protect against a patient’s threatened violent behavior except where the patient has communicated to the psychotherapist a serious threat of physical violence against a reasonably identifiable victim, and that the duty is discharged by making reasonable efforts to communicate the threat to the victim and to a law enforcement agency. Ewing v. Goldstein (2004) 120 Cal.App.4th 807 held that a threat relayed by a patient’s immediate family member, shared to further the patient’s treatment, can count as a patient communication. Whether the statute was triggered on a given set of facts, and whether the response was reasonable, is where these cases are litigated.
What makes a risk report hold up on cross-examination?
Traceability and restraint. Every conclusion should be followed to the record that supports it, every instrument should be justified for this examinee and this question, and the limits of the method should appear in the report rather than being extracted by opposing counsel.
The predictable failure modes are the opposite of that: an opinion resting on interview alone, an actuarial percentage attached to one person, a psychopathy score introduced where psychopathy was never the issue, an instrument used on a population it was not validated for, and a report with no statement of what would lower the risk. A risk assessment cannot tell anyone what a specific person will do. It can say what the relevant factors are, how strong the evidence behind each one is, and what would change the picture.
Send the question, not just the file.
Tell us what decision the risk opinion has to support and by when, and you will get a straight answer on whether the available record can carry it.