The error detection test lives on a single ugly fact: the characters that matter most look almost identical. O and 0, I and l, 6 and 9. Three sub-tasks probe the same skill from different angles: comparing callsign pairs, judging two callsigns for identity, and finding the one odd character hidden in a dense grid. This guide covers all three formats, the confusable families, and the scanning discipline that catches single-character errors on a clock.
What is the Error Detection Test?
The test measures perceptual precision at speed: reading strings exactly as written, not as expected. Expectation is the enemy; the brain autocompletes THY4KL and THY4KI into the same word unless forced to check character by character. Aviation punishes that autocomplete brutally. A controller confusing two similar callsigns issues a clearance to the wrong aircraft; a pilot misreading one digit in a squawk code triggers a phone call nobody wants.
The three sub-tasks rotate within one session. Pair counting shows four callsign pairs and asks how many match exactly. Identity judgment shows a reference and a copy and asks a plain yes or no. Visual search fills a grid with one repeated character and hides a single different one in it. All three run against per-question timers that tighten with level, and its close cousin in the pilot battery is the perceptual speed test, which extracts values instead of differences.
Which Exams Include the Error Detection Test?
Character-level checking tasks appear across controller and airline screening. Candidates commonly report this test type in:
- FEAST and related ATC selection batteries, where strip checking is a native job skill
- DLR screenings, in perception and accuracy stages
- Aon / cut-e batteries, in comparison and checking scales
- ADAPT, within its perceptual tasks
- COMPASS and Advanced COMPASS by EPST, in speed and accuracy sections
Providers change their content without notice and publish no official syllabus, so treat this as candidate-reported guidance.
How the Test Works: Format and Rules
Three question types share the session:
- Identity judgment: a reference callsign and a copy; answer Yes if identical, No if anything differs.
- Pair counting: four callsign pairs; count how many are exact matches and answer the number.
- Visual search: a grid of one repeated character with a single stranger among them; tap the stranger.
- Differences are always single-character confusables: O against 0, I against l, 5 against S, 8 against B, 6 against 9.
- A timer bar drains per question and turns red near the end; wrong answers reveal the correct one.
Suppose the reference reads THY4KL and the copy reads THY4KI. Five characters agree, the last does not: the answer is No. This is exactly how it looks in the test:
The search grid and the character families it is built from:
The level ramp raises string length and tightens the clock: Level 1 gives 20 seconds per question with three-digit flight numbers, Level 2 drops to 15 seconds and adds a fourth digit and the visual search task, and later levels compress further while adding suffix letters.
Your brain reads words. This test pays only for reading characters.
Scoring: What a Good Result Looks Like
Accuracy per question with response time recorded. The distractor logic makes error types diagnostic: a No answered on an identical pair means over-suspicion, a Yes on a confusable pair means autocomplete won. Strong candidates run ninety percent plus with the errors, if any, on the tightest timers rather than on the classic confusables. Missing an O versus 0 difference at a comfortable time limit is the result assessors remember.
How to Prepare: Strategies That Actually Work
- Scan character by character, left to right. Pace it like a metronome. The only reliable defense against autocomplete is serial checking.
- Learn the confusable families cold. O0, Il1, 5S, 8B, 69, 2Z. When you know where errors can live, your eye slows exactly there and nowhere else.
- Compare in vertical slices on pair tasks. First character against first character, second against second. Never read each string whole and compare impressions.
- On search grids, scan rows serpentine-style. A structured sweep finds the stranger reliably; random darting finds it by luck.
- Call differences out loud in practice. Kilo versus India. Naming the mismatch trains the discrimination deeper than silent spotting.
- Warm up your value-reading too. The perceptual speed test trains the same eye at the same tempo on a different target.
Common Mistakes to Avoid
- Reading callsigns as words. THY4KL is six characters, not one lump.
- Declaring identity after checking only the prefix, where airlines rarely differ.
- Slowing globally instead of locally; spend your caution on confusable positions, not on every character equally.
- Forgetting that suffix letters at higher levels are where the single change loves to hide.
Practice This Skill
Character-level precision is a habit built at speed. Train with the Error Detection module, which reproduces the three sub-tasks, the confusable families and the tightening timers exactly as described here. Level 1 is free, and the guided tutorial walks you through the format before your first attempt.