Prepare for the NCRA Certified Broadcast Captioner exam by practicing decision routines under realistic audio: how you respond to mistranslates and untranslates, how you attribute speakers, and how you hold a stable delay behind the feed. Use the scenarios, the error-response table, and the self-scoring rubric here as a weekly cycle, and confirm all administrative exam details directly with NCRA.
When a stroke goes wrong live: rehearsed recovery beats hesitation
The demanding part of realtime captioning is not writing fast; it is deciding, within a second or two, whether an error is worth fixing. Rehearsed routines for mistranslates, untranslates, and misstrokes keep that decision quick and consistent.
Trace a concrete scenario: mid-weather report, you write 'their' for 'there.' A workable routine is to correct only errors that change meaning — 'site' versus 'cite' matters; 'their' versus 'there' in a forecast usually does not. The plausible mistake is treating every homophone as urgent, which costs attention and widens the gap to the audio. Recovery routines matter because delay compounds: every second spent polishing a low-impact word is a second you cannot spend on the next sentence.
Learn the distinction between an untranslate and a mistranslate, because they signal different problems. An untranslate is steno that surfaces as raw strokes or asterisked text — a missing brief. A mistranslate is a real but wrong word — usually a brief conflict or a stroke slip. Untranslates are best logged for dictionary work after the shift; mistranslates deserve a same-shift check for conflict. Deciding which type you just produced, almost automatically, is the skill to drill rather than the individual words themselves.
| On-air error type | Immediate response | Post-shift fix |
|---|---|---|
| Mistranslate that changes meaning | Correct immediately, even at a cost in lag | Check the conflicting brief against practice dictation |
| Mistranslate that does not change meaning | Let it stand, keep writing | Log it for the next dictionary maintenance session |
| Untranslate (raw strokes appear) | Fingerspell a short phonetic version if the word is key; otherwise move on | Write and test a new brief in drill audio before trusting it live |
| Fingerspelling overrun | Stop the finger letters cleanly and resume with the next phrase | Shorten the routine you use for long or unfamiliar names |
| Dropped sentence after a stall | Resume at the current spoken word; do not backfill | Identify the trigger passage type and drill it specifically |
Homophones, numbers, and proper nouns: dictionary habits that hold up live
Dictionary quality determines how much recovery work you ever face. Build briefs for names and local terms, write numbers with explicit formats, and audit conflicts in drill audio before they surface during a broadcast.
Consider a realistic scenario: a regional news story suddenly needs 'Chattanooga' repeatedly, and your only brief for those strokes collides with a common word. The plausible mistake is adding a new brief mid-broadcast without testing it, risking a fresh conflict in the next sentence. The better decision is to fingerspell a shortened phonetic version on air, then add the brief and verify it against dictation the same day. New briefs earn live trust only after they survive practice audio, which is why dictionary maintenance and writing drills are one activity, not two.
Numbers deserve deliberate stroke habits because they arrive as dates, money, scores, and percentages in the same passage. Decide in advance how you write each class — a score as digits, a quoted amount with its unit, a year without an ordinal — and keep those habits consistent across program types. A concrete exercise: caption a 60-second sports recap from audio, then compare your number strings against the transcript word by word. Watch specifically for digits transposed, currencies attached to the wrong figure, and percentages written as plain numbers.
Speaker identification and multi-voice panels: marking changes without falling behind
In multi-speaker audio you must attribute lines and punctuate turn changes while still writing the next sentence. Practical conventions — labels, dashes for interruptions — need to be automatic habits, not inventions made on the spot.
Picture a two-minute exchange where two debaters interrupt each other and the moderator interjects. The plausible mistake is pausing to confirm who spoke before writing anything, which forfeits whole sentences. The better decision is a standing convention: mark every turn change with your chosen symbol or label the moment the voice shifts, and resolve ambiguous attributions at natural pauses when the video or context confirms who is talking. Because attribution is partly inference under time pressure, your convention should have a defined fallback — an unnamed speaker tag — rather than silence.
Offline captioning of the same panel hides this difficulty, since you can rewind and verify each speaker at leisure; live work does not. To train the live skill, take a recorded panel with at least three voices, caption two minutes straight through without stopping, then compare your attributions and interruption marks against the video. Expected observations: your errors cluster at fast crosstalk and at voice changes that lack visual cues. That tells you which situations to drill with deliberately chosen audio, rather than re-captioning the same easy material.
Latency and pacing: writing behind the feed on purpose
Live captions trail the audio by a delay that varies with the setup. Pace your writing so your personal lag stays stable, and use natural program pauses to close gaps instead of rushing stroke-for-stroke at every sentence.
Distinguish three things that are easy to conflate: broadcast delay built into the transmission chain, processing time in the captioning system, and your own writing lag behind what you hear. In many live setups, captions arrive a few seconds after speech, and your controllable share is the writing lag. The plausible mistake is writing at maximum speed from the first word, then stalling entirely when a dense passage arrives, so the gap swings wildly. The better decision is an even, sustainable pace that treats lag as a budget you keep roughly level, spending surplus attention on hard passages rather than banked speed.
A practical drill makes lag visible. Choose a recorded newscast, caption three minutes, and note the time when a given sentence was spoken and when you finished writing it, sampling at the start, middle, and end. Expected observations: lag grows after numbers, names, or crosstalk and shrinks during slow segments, which shows you where your pacing breaks down. Repeat with different program types and you will typically find the same pattern concentrated around your weakest content categories — that concentration, not the average lag, is what to work on next.
Verbatim or clean read: matching your output to the program
Not every program needs word-for-word captions. Decide in advance which register each program type requires, because switching strategy mid-broadcast costs far more attention than committing to either approach alone.
Verbatim captioning reproduces every spoken word, including false starts and fillers; a clean read conveys meaning while trimming repetitions and verbal debris. Trace a scenario: a press conference where the speaker restarts sentences repeatedly. The plausible mistake is defaulting to full verbatim out of caution, capturing every restart and losing the thread of the next answer. The better decision is a pre-decided threshold — keep substantive content, drop pure restarts and fillers — applied consistently, so each false start takes a fraction of a second of judgment instead of a full sentence of writing.
The register decision belongs to the program and any stated requirements, not to personal preference: proceedings and scripted content generally call for closer to verbatim, while fast unscripted talk tolerates a cleaner read. Where a requirement is stated for the material you caption, that requirement governs your output. The training implication is to drill both registers deliberately — caption the same audio once verbatim and once as a clean read, then compare which decisions you made differently. This comparison also reveals which trim decisions you make inconsistently, which is exactly where a pre-set threshold helps.
A self-scoring drill: measuring accuracy and untranslate management
Score your own captioning against a fixed rubric instead of a vague sense of how a session felt. Percentage accuracy, untranslate counts, and latency drift together give you a weekly trend line you can act on.
Run the drill the same way each week. Take a ten-minute recorded news segment you have not heard, caption it in one pass without pausing, then produce a score by comparing your output to a clean transcript: count word errors (wrong words, omitted words, untranslates, wrong numbers) against total words for a percentage. Separately, log untranslates and mistranslates in two lists, and note any point where you dropped a full sentence. The separation matters because the two error types have different fixes — dictionary additions for untranslates, conflict audits and stroke review for mistranslates — and a single blended error count hides which fix is needed.
Interpret the results as a trend, not a verdict. A useful self-check rubric: across consecutive weeks, word-error percentage trending down, untranslates migrating from common words to genuinely rare terms, and no sentences dropped during the last third of the session. When a number stalls, change the drill rather than the goal — for example, swap in audio with heavier name density if fingerspelling errors dominate. Treat these self-check scores strictly as training milestones for your own practice; they indicate where to focus, not any prediction about exam performance or a passing threshold.
- Weekly drill: one unseen ten-minute segment, one pass, no pausing.
- Score word errors as a percentage against a clean transcript.
- Log untranslates and mistranslates in two separate lists for two different fixes.
- Mark every dropped sentence and identify the trigger: numbers, names, crosstalk, or pacing.
- Compare against last week before choosing the next drill, and change only one variable at a time.
A preparation sequence and readiness checks before test day
A workable sequence runs baseline testing, dictionary maintenance, daily dictated practice, weekly self-scoring, and full mock sessions. Readiness shows up as trend direction and recovery speed, not as one unusually good run.
An adaptable sequence: in week one, run the self-scoring drill twice to set a baseline and audit your dictionary for conflicts using drill audio. In the following weeks, caption something daily at your current sustainable speed, hold one longer session weekly, and spend one session maintaining briefs — adding, testing, and pruning — rather than writing. Every third week, run a full mock: multi-speaker audio, numbers and names dense, captioned start to finish, then scored. Adjust the program mix to whatever you will actually caption, and scale session length to your schedule rather than copying someone else's volume.
Concrete readiness checks: your word-error percentage has fallen over several consecutive drill weeks; you recover from a deliberate bad stroke without dropping the following sentence; your lag stays within a narrow band across a whole segment; and your untranslate list has shifted toward genuinely obscure terms. If any check fails, the fix is targeted drilling of that failure mode, not more general writing. For administrative questions about the credential itself — eligibility, scheduling, and current requirements — NCRA is the authority; its site at ncra.org carries the official information, and this guide deliberately leaves those logistics to the issuer.
References and further reading
Use these references to explore the concepts and check the latest information from the relevant organizations.