Spelling Out Loud in English: You Say "E," They Write "A"

You spell your last name over the phone, hang up satisfied, and the confirmation email comes back with two letters swapped. The line was fine. English and Spanish letter names overlap exactly where your ear draws no line: say e the Spanish way and you produce a vowel that points straight at English A; say i and you point at E. The alphabet is the first thing you learn and the last thing anyone corrects, because nobody spells out loud in class and everybody spells out loud in real life. Talking slower won't fix that. Knowing which four pairs collide will, and so will trading them for anchor words. Aviation and call centers hit the same wall and landed on the same answer.

Your "e" lands on somebody else's A#

The collision is about vowels, not about your accent. English names A as /eɪ/, which sits right next to the Spanish e. It names E as /iː/, right next to your i. And it names I as /aɪ/, which sounds like nothing you would ever produce. What you get is a chain shift: your e gets written down as A, your i as E, and your a, a short clean vowel English uses for no letter at all, comes out sounding like R to an American ear, which calls that letter /ɑr/.

The mismatch underneath has been measured. English runs twelve pure vowels and eight diphthongs against Spanish's five pure vowels, so several English vowels crowd into one mental slot in your head. Norman Coe wrote the sentence that still describes the problem best back in 2001, quoted by Uribe Enciso and colleagues: "Typically, at least two English vowels share the 'phonetic space' occupied by one Spanish vowel, so one-to-one correspondences are practically impossible."

Eight letters, one single vowel#

Eight letters of the English alphabet carry the exact same vowel sound, /iː/. Jonathan Marks lists them flatly in Cambridge's English Pronunciation in Use: "Eight letters of the alphabet have the sound /iː/... B C D E G P T V." Nothing separates those eight but the opening consonant, and the opening consonant is the first thing a phone line, an airport PA, or a café with music takes away from you.

This is not a personal failing. Machines get it wrong too. Ronald Cole and Mark Fanty opened their work on spoken-letter recognition by calling it one of the hardest tasks in automatic speech recognition, and they said why in one line: "The difficulty of the task is due to the acoustic similarity of many of the letters." The pairs their system had to untangle (B/D, B/P, D/T, T/G, C/Z, V/Z, M/N and J/K) are the same ones that wreck your zip code.

G, J and Y: the same trap in a different costume#

G and J are the same English consonant, /dʒ/, split only by the vowel that follows it: gee /dʒiː/ against jay /dʒeɪ/. The G/J pair is the E/A pair again in a new costume. Spanish leaves you nothing to fall back on here, since it calls g /xe/ and j jota, so what comes out of your mouth is neither one. The industry has this on file. The list of easily confused letters at Call Centre Helper, written by Jonty Pearce, includes an entry that reads like it was written for this article: "G and J (particularly when pronounced by a non-native speaker)." The same list adds "Y and I", and Spanish pushes you the wrong way there too. Calling Y i griega reinforces the exact association English needs you to break, because wye /waɪ/ and I /aɪ/ already sound too close without the help.

Spanish had to solve its own version of this, which is oddly reassuring. Spain's Royal Academy now recommends calling Y ye, and it says why: the choice "is justified by its simplicity, since it differs from the name of the letter i without needing any qualifier". It's the same academy that documents why Spanish speakers say be larga and ve corta: be and ve are pronounced exactly the same, and without a surname there's no way to tell them apart by ear.

What one misheard letter actually costs#

A swapped letter isn't cosmetic. It's different data. Call Centre Helper makes the point with a British postcode: without spelling the letters out, "NP2 3BP" can easily be heard as "MB2 3PP", an address that doesn't exist. Banking, insurance and healthcare teach the spelling alphabet on day one of training for exactly this reason.

In healthcare the price goes up. Jason S. Adelman, patient safety officer at Montefiore, opens his commentary for the US federal agency AHRQ by stating that "wrong patient errors are omnipresent in health care." The experiment he cites is the frightening part: with eye tracking on 25 clinicians, not one of them checked the second identifier when picking a patient from a list that held both Jessie Torres and Jessica Torres. A well-spelled last name and a half-heard first name are enough to land you in the wrong chart.

Automatic recognition won't rescue you either. A PNAS study of five commercial speech recognition systems (Amazon, Apple, Google, IBM and Microsoft) measured a word error rate of 0.35 for Black speakers against 0.19 for white speakers, and the gap held just as wide on identical phrases spoken by both groups. That study didn't test Spanish-accented English. What it establishes is the thing that matters here: recognizers are not neutral about the voice they hear. They inherit the mold they were trained on.

Aviation already solved this for you#

The right answer was invented seventy years ago and it's called a spelling alphabet. ICAO says so plainly in its own history: some letters sound alike, "M and N or G and J", and that breeds confusion between people with different accents or over a poor line. The organization tested candidate words with speakers from 31 countries before locking the final version on March 1, 1956, the same day it took effect across NATO. One detail works in your favor: the words were picked to function in English, French and Spanish at once, so Alfa, Bravo and Charlie are already built for your mouth.

In practice you don't need all 26. Memorize the six that fail you (Alfa, Echo, India, Golf, Juliett, Yankee) and deploy them only when they're due: "E as in Echo." Then add two conventions English takes for granted and Spanish doesn't. Repeated letters get double, which the British Council teaches at A1 level by spelling Reeves as "R-double E-V-E-S". And Z is zee in the US, zed in the UK. Could you spell your full last name out loud right now without stalling on the E or the I? That gap between recognizing a vowel and producing it under pressure is what Tabla drills: you do the talking, a speech recognizer scores what comes out, and the sound you miss comes back around later.

You're giving your email over the phone in Chicago and you have to spell "gema." What do you say, and what happens if you don't?

You say "G as in Golf, E as in Echo, M as in Mike, A as in Alfa." Skip the anchors and your G lands as /dʒiː/, competing with J /dʒeɪ/; your Spanish-style e lands on A; your a gets written as R. The plausible result is jamr or jama. All four letters of that one short word sit inside the collision pairs, which is why you start anchoring at the first letter instead of halfway through.

The next time someone asks "Can you spell that?", don't speed up. Slow down on E, I, G and A, and anchor those four with their word. The rest of the alphabet takes care of itself.

Try this today: record yourself spelling your last name and your email twice, first plain and then with anchors ("E as in Echo"), and check whether you could write it down correctly from your own recording.