Prepare for the NVRA Certified Verbatim Reporter (CVR) exam by treating every dictation drill as a decision exercise, not just a speed test. For each practice passage, apply four rules: write exactly what was said without smoothing it into cleaner English, punctuate by grammatical meaning rather than breath pauses, attribute every speaker change explicitly, and flag anything uncertain instead of guessing. Score each drill with a five-part rubric and re-drill only the categories where you lose points.
What "verbatim" obligates you to write, and what it does not
Verbatim capture means recording every word, false start, and correction exactly as spoken, rather than producing a cleaned-up, readable summary. The distinction is the foundation of the CVR skill set, so your drills should train faithful capture before they train speed.
Contrast the two modes directly. A witness says, "I—I left around, uh, six-ish, no wait, closer to seven." Clean transcription would compress this to "I left around seven." True verbatim preserves the stutter, the filled pause or its agreed marker, and the self-correction. Practicing on polished news audio hides this problem; practicing on natural, unscripted speech, with interruptions and restarts, exposes whether you actually write what you hear.
Verbatim does not mean writing sound effects, off-microphone murmurs you cannot resolve, or reorganizing sentences for clarity. It means the words spoken by the speaker of record are all present and unaltered. In your drills, mark every place where you silently fixed a speaker's grammar or dropped a repeated word. Those marks are your training targets, because smoothing is the habit most at odds with the verbatim standard.
Conflict triage: flag first, fix deliberately, never guess a name
Whether you write by voice recognition or steno, your dictionary produces wrong-word conflicts under speed. Train a fixed triage order: verify structure, flag anything unverifiable, and correct flagged items immediately after the passage while context is fresh.
Worked scenario: dictating a testimony passage, a voice writer says "decedent's prior testimony" and the software renders "the sedans prior testimony." The plausible mistake is to keep dictating, planning to fix everything at the end. Untranslated or mistranslated words pile up, edit time balloons, and under live conditions the text stream becomes useless. The better decision is a two-step habit: keep a minimal placeholder flow so you do not lose the speaker, then correct the conflict as soon as the sentence ends, and afterward add a dictionary entry for "decedent's" so the conflict never returns. Why it matters: an uncorrected wrong word is a substantive error in the record, and a collision you patch repeatedly is proof your dictionary needs maintenance, not more proofreading.
Proper nouns deserve a separate rule: never resolve an unfamiliar name from sound alone. "Carlson" and "Karlsen" are equally plausible. In a drill, write your best phonetic form, mark it as unverified, and check it against any supplied name list afterward. Build this into speed practice from day one, because a flagging habit formed at slow speeds is the one that survives when the dictation gets fast.
Punctuation for meaning: commas that change who is bound by a clause
Punctuation on a verbatim record is a meaning decision, not a breathing decision. Train yourself to identify sentence structure first, then punctuate, then read the sentence back aloud to check that a listener would draw the intended conclusion.
Worked scenario: a speaker dictates, "All deponents, who have reviewed the exhibit, must sign the certification." With the commas, the clause is nonrestrictive: every deponent must sign, and the reviewing is incidental background. Without the commas — "All deponents who have reviewed the exhibit must sign the certification" — the clause is restrictive, and only deponents who reviewed the exhibit must sign. The plausible mistake is punctuating where the speaker pauses, which reflects the speaker's lungs rather than the sentence's logic. The better decision is to parse the structure first, ask who the sentence actually binds, choose the punctuation that produces that meaning, and confirm with a read-back.
Make this a scored drill component. Take five complex sentences from any instructional or legal-style text you did not write, punctuate them cold, then compare against the source. For each disagreement, classify it: same meaning, different meaning, or unreadable. Only the middle category is a genuine error to fix; the last tells you to simplify your sentence-structure parsing before adding speed. Expect early drills to show rhythm-based punctuation dominating your errors, which is exactly the habit the read-back check is designed to break.
Speaker attribution in multi-party testimony
Every speaker change must be attributed correctly, even when speakers interrupt, finish each other's sentences, or answer a question you did not hear clearly. Train attribution as its own layer, separate from word capture.
The failure mode looks like this: an examiner asks a question, a second party interjects "she already answered that," and the witness continues. A writer focused on the words may assign the interjection to the witness, producing a transcript where the wrong person makes the statement. The better decision is to treat the first word of every utterance as an attribution checkpoint: before capturing the content, decide who is on the record. When attribution is genuinely unclear, mark it for follow-up rather than assigning it silently.
Drill this with two- and three-voice audio, deliberately chosen from material with interruptions. First pass: capture words only. Second pass: replay and fix speaker labels only. Comparing the passes shows you the attribution errors your word-focus hid. Over successive drills, the gap between the passes should shrink, which tells you attribution is becoming automatic rather than a layer you add afterward. If the gap never shrinks, slow the material down; attribution speed is built on habit, not effort.
Numbers, spoken arithmetic, and figures under speed
Spoken numbers are the highest-consequence verbatim material because a dropped digit silently changes the record. Train a fixed conversion rule for how you render spoken numbers, dates, and amounts, and apply it identically in every drill.
A speaker says "one hundred forty-two thousand, five hundred and six dollars." Under speed, writers commonly drop a magnifier ("forty-two thousand" becomes "forty-two"), merge a correction ("it was five—no, six hundred"), or transcribe the amount in a form that no longer shows how it was spoken. Choose one convention, such as rendering numerals for amounts while preserving a spoken correction in full, apply it uniformly, and drill dictation packed with amounts, dates, phone-style sequences, and comparisons. The goal is not style elegance; it is that every figure on your page is reconstructable to what was said.
Check yourself with a two-column method: after transcribing, write the numbers you captured in one column and re-listen to only the number-bearing sentences, listing what was said in the other. Mismatches here are disproportionately serious relative to their count, because one wrong digit changes a total. Expect your first attempts to reveal dropped magnifiers and lost self-corrections; those two patterns should be your explicit targets in the following week of drills.
A scored dictation drill with a self-check rubric
Use one repeatable drill: two minutes of unscripted two-voice audio, transcribed cold, then scored against a five-part rubric. The rubric converts vague "accuracy" into categories you can retrain individually.
The drill: source any two-minute recording of natural, unscripted speech with at least two speakers (public archival audio or a study partner reading prepared material with interruptions). Transcribe once, without replaying, under whatever timing you can currently sustain. Then score it with the rubric below, re-listening only to the sentences where you suspect an error. Record your five scores with the date. Scores are learning milestones for your own trend line, not a prediction of any test result.
Expected observations on early drills: dropped function words ("a," "of," "that") cluster around fast passages; proper nouns get guessed rather than flagged; punctuation follows breath; self-corrections get silently resolved. After two weeks of targeted re-drills on your lowest category, re-score. A useful milestone is your lowest category rising by one rubric point while your total climbs steadily. If one category stays flat across three drills, change the drill for that category rather than adding volume.
| Rubric category | Score 0 | Score 1 | Score 2 |
|---|---|---|---|
| Verbatim fidelity | Repeated dropped or smoothed words change sense | Isolated small-word drops, meaning intact | Every spoken word and correction captured |
| Punctuation for meaning | A reader would misread an obligation or speaker's point | Ambiguous but not misleading | Read-back matches intended meaning |
| Speaker attribution | At least one statement assigned to the wrong speaker | All attributed; one change marked uncertain | All changes attributed confidently and correctly |
| Numbers and names | A figure changed or a name invented | Numbers intact; names flagged not guessed | All figures exact; all names resolved or flagged |
| Flag discipline | Uncertainties silently resolved | Some flagged, some guessed | Every genuine uncertainty flagged and later resolved |
An adaptable preparation sequence and readiness checks
Sequence your preparation in rotating weekly blocks: capture fidelity, conflicts and dictionary work, punctuation, attribution, then a mixed mock drill. End each cycle with the rubric, and move on only when a category holds its score.
A four-week adaptable rotation: Week 1, fidelity — daily drills on unscripted speech with the fidelity category scored; add a running list of words your dictionary or fingers get wrong. Week 2, conflicts — every drill ends with dictionary maintenance for that session's conflicts and a name-flag review. Week 3, punctuation and attribution — alternate days, using the two-pass attribution method and the five-sentence punctuation check. Week 4, mixed — full two-minute drills scored on all five categories, with any category scoring below your Week 3 result re-drilled the following week. Compress or extend the rotation to fit your available weeks; the order matters more than the duration.
Readiness checks before any formal testing: you can transcribe unfamiliar unscripted audio at a sustained pace without silently smoothing the speaker's words; your conflict list from three consecutive drills is shrinking, not growing; your punctuation disagreements with source texts are same-meaning only; and your rubric totals are stable across two consecutive drills. Administrative matters such as scheduling, eligibility, and the exact formats and conventions used in the current exams are set by the association itself — confirm those directly through the National Verbatim Reporters Association at nvra.org rather than relying on secondary descriptions. When all checks hold, shift your remaining time from new drills to maintaining dictionary and rhythm consistency.
References and further reading
Use these references to explore the concepts and check the latest information from the relevant organizations.