Voice Stress Analysis: How It Works, Tests & Limitations

What is CVSA? VSA, AISIS, and VASIS explained

CVSA stands for Computer Voice Stress Analyzer, a commercial product offered by NITV Federal Services. Voice stress analysis is the broader field. AISIS and VASIS are separate VIPRE products. A study should identify the product, version, task, and procedure it evaluated before its results are applied to another system.

A difference in product design is not, by itself, evidence of better performance. Likewise, research on speech under stress is not interchangeable with research on detecting deception. Each claim needs evidence suited to the outcome being claimed.

What is voice stress analysis?

The term covers approaches to examining how stress may affect the voice. Research on speech under stress and the use of commercial systems in investigative interviews are related subjects, but they answer different questions.

A voice stress analysis test is one structured application: an examiner prepares questions, records spoken answers, and reviews indications in context. Evidence that stress can change speech does not, by itself, show that a system can distinguish truth from deception.

What is a voice stress analysis test?

A voice stress analyzer examines features of a person’s speech during an interview. It may help a trained examiner identify responses that deserve a closer look. It does not read a person’s mind or establish that a statement is true or false.

VIPRE offers AISIS and VASIS for trained investigative professionals. The examiner asks structured questions, reviews the responses in context, and decides what follow-up is warranted. A result is an investigative lead, not a finding of guilt or proof of deception.

How voice stress analysis works

A microphone captures a person’s answers. Software displays or analyzes aspects of the voice so an examiner can review changes across a structured examination. The exact measures and workflow depend on the system. AISIS adds AI-assisted review while leaving the trained professional in control. VASIS provides an industry-standard voice stress analysis platform.

Stress can affect speech, but it can have many causes, including fear, confusion, fatigue, illness, trauma, or the pressure of being questioned. An observed voice change cannot, by itself, tell an investigator why it occurred. Question design, examiner training, context, and independent corroboration remain essential.

What happens during a voice stress analysis test?

The examination combines preparation, recorded questions and answers, and examiner review. A microphone captures spoken responses; the system supports analysis of voice patterns. The exact sequence depends on the system, examiner training, and agency procedure.

Before recording, the examiner establishes the purpose of the examination, explains the procedure, and clarifies the questions so the person understands what is being asked. During the examination, the examiner follows the applicable protocol and considers recording quality and interruptions.

Afterward, the examiner reviews the responses and relevant circumstances, documents observations, and identifies areas requiring clarification or independent investigation. A software indication does not replace that professional review.

Where it may fit in an investigation

An agency may consider voice stress analysis for a structured interview, preliminary screening, or the review of areas that need clarification. It is most useful when the examiner can follow up with documents, physical evidence, witness accounts, or another independent source.

Before selecting a system, agencies should ask who will administer the examination, what training is required, what the output actually says, whether the procedure can be documented and reviewed, and how an indicated response will be checked against outside evidence.

What an agency should document

The examination record should let another authorized reviewer understand what was asked, what was observed, and what happened next. Preserve the relationship between the recording and the examiner’s notes so later decisions can be traced to the actual exchange.

• The examination purpose and the question set used
• The recorded responses and any recording-quality concerns
• Clarifications, interruptions, and relevant interview circumstances
• The examiner’s observations and their basis
• The independent follow-up assigned and what it established

Voice stress analysis test questions

Questions should be clear, focused, and appropriate to the examination’s purpose. A trained examiner develops the question set under the applicable protocol, using defined events and time periods so an answer can be understood in context.

For an agency reviewing question quality, compare a broad question such as ‘Have you ever done anything dishonest?’ with a focused question such as ‘Did you remove that envelope from the records room on Monday?’ The second identifies an action, an item, a place, and a time. This is an illustration of clear wording, not a validated test format or a ready-to-use examination.

The examiner should resolve ambiguous terms before recording and avoid combining several separate allegations into one question. A list found online cannot replace certification, case preparation, or the procedure required for the system being used.

• Define the specific issue and relevant time period
• Use wording the person understands
• Keep separate issues in separate questions
• Document the questions actually asked and any clarifications

What do voice stress analysis test results mean?

A voice stress indication identifies a response for closer examiner review. It does not establish why a person was stressed, whether a statement was false, or whether an allegation is proved. The examiner considers the question, the recorded answer, the interview context, and information available from other sources.

No instrument directly measures a lie. Voice stress systems assess features of speech that may relate to stress, while a polygraph records several bodily responses under a different examination procedure. An investigative lead and a verified finding are different outcomes.

A useful examination record connects each observation with the question asked and the follow-up required. Investigators can then check the issue against documents, witness accounts, or other independent evidence rather than treating an indication as a final decision.

What the research shows

Research on speech under stress does not automatically validate a device for detecting deception. A well-known National Institute of Justice field study found weak deception detection for CVSA and LVA in interviews about recent drug use. It did not test VIPRE’s VASIS or AISIS, so its result should not be presented as a direct test of those products.

The absence of a direct test is also not evidence that a different product has a particular accuracy rate. Agencies should ask for product-specific evidence, including the study design, population, ground truth, missed deceptive answers, false alarms, and inconclusive outcomes.

Evaluate a system on its evidence and workflow

A field demonstration can show usability. It cannot establish an accuracy rate. Agencies should assess the operating procedure, training requirements, validation evidence for the exact product and version, definitions of reported outcomes, and examples of limitations or inconclusive results.

A responsible evaluation also considers whether the system preserves an examination record, supports an examiner’s review, and fits the agency’s process for documentation and corroboration. VIPRE can provide VASIS and AISIS workflow and training information for agencies assessing their operating environment.

VIPRE experience and examiner preparation

VIPRE Technology Group provides AISIS and VASIS systems, examiner certification, and operational support. Its team includes advanced certified voice stress analysis examiners, police academy instructors, and professionals with investigative and military intelligence backgrounds.

VIPRE is the manufacturer describing its own systems in this guide. Product knowledge and examiner training help explain the workflow; they do not replace independent validation or establish an accuracy rate.

What newer voice research contributes

A 2025 systematic review by Schewski and colleagues examined acoustic correlates of negative emotions, stress, and cognitive load. Some voice features showed potential as indicators, while study quality, individual differences, and limited real-world research constrained the conclusions.

The review expands the research context beyond older commercial deception tests. It does not establish an accuracy rate for AISIS or VASIS, validate an inference of deception, or show that an audio recording directly measures vagal activity.

Research references and scope

These sources address different research questions. Read each study in relation to the systems, participants, and procedures it actually examined.

Van Puyvelde and colleagues (2018) review voice production under physiological, cognitive, and emotional load. Their framework provides background on voice and stress, not independent performance validation of AISIS or VASIS.

Damphousse and colleagues (2007), in an NIJ-sponsored jail study, evaluated CVSA and LVA using statements about recent drug use. Detection did not exceed chance in that setting. The study did not evaluate AISIS or VASIS; that distinction does not validate either VIPRE product.