Korean stops come in three series and English has two, which is the harder half of this subject and the one nobody warns about. English separates its stops by voicing, so b and p differ in whether the vocal cords are running. Korean separates its by how much breath and tension there is, in three steps, and none of the three lines up with an English b or an English p.
So an English ear arrives with two boxes for three sounds and files two of the Korean series into the same one. That is a category problem rather than a muscle problem, and no amount of repeating the sound aloud will fix it.
Six things to put in your mouth
- ㄱ, ㅋ and ㄲ are three different sounds, plain, aspirated and tense. English has two of the three and no speaker controls the third without practice, and the distinction separates real words.
- final consonants are unreleased, stopped without letting the air go, so a syllable ending in k closes rather than pops.
- 한국말 assimilates: the k at the end of one block becomes ng before the m of the next, so it is said hangungmal. Korean is full of these changes and they are regular.
- ㅓ and ㅗ are different vowels that both get written as o in romanisation, which is why romanised Korean is unreliable and hangul is not.
- there is no f, v or z, so loanwords are reshaped: coffee arrives with a p sound.
- ㄹ sits between l and r, closer to a tap between vowels and to an l at the end of a syllable, and it is one letter doing both jobs.
The spelling keeps the word and the rules reshape it
The other half is that Korean writes what a word is rather than what it sounds like in the sentence it is sitting in, and then applies a set of rules at every boundary. A final consonant slides onto the next block when that block begins with a vowel. Consonants assimilate to their neighbours, so a k before an m comes out as an ng. An h next to certain consonants merges into them and disappears.
That is a good design decision and it has a two sided cost. You cannot pronounce a word correctly from its spelling until you know the rules, and you cannot spell a word correctly from hearing it, because the spelling has preserved a form the pronunciation hid. Both failures come from the same choice, and both are repaired by the same fortnight of work on the rules themselves.
The rules are finite and they are not many, which is the good news. Six or seven of them account for almost everything you will meet, they can be listed on one page, and once they are automatic the spelling stops being a trap and becomes what it was designed to be, which is a way of keeping a word recognisable across all its forms.
Two series heard as one, for years
Here is the stall. A learner hears the plain series and the tense series as the same sound, because both are voiceless to an English ear and the difference is tension rather than voicing. They then produce one sound for both, are understood almost every time because context repairs it, and never receive the correction that would split the categories.
The specific damage is on the receiving side rather than the producing one. Two words differing only in that contrast arrive as the same word, so the listening has a permanent hole in it that looks like a vocabulary gap. The exit is minimal pairs, listening only, before any production practice: play the pair, choose, check, and do it a few hundred times over a fortnight. Working the ear before the mouth is the general rule for any sound your first language does not have, and it is argued out in the stalls every learner hits, and the way through each.
One rail, three series
Where the three Korean stops sit, and where English puts its two
The rail is breath and tension rather than voicing, which is the axis English does not use, and the two English sounds land in the wrong places on it.
Placed by how much breath and tension the sound carries. The English positions are approximate and are there to show the mismatch rather than to be copied.
- The tense seriesTight, short, with almost no air after the release
- English bVoiced, which is an axis Korean is not using here
- The plain seriesBetween the other two, and voiced between vowels
- English pAspirated at the start of a word, and not elsewhere
- The aspirated seriesA clear puff of air, and the easiest of the three for an English speaker
The two English marks are why this is hard rather than merely new. An English speaker does not have a spare category to put the third Korean series in, so it lands on top of one of the other two and stays there.
Two ways to see a distinction you cannot hear
Praat, the free acoustic program from the University of Amsterdam, is unusually well suited to this particular problem. What separates the three Korean series is largely the delay between releasing the consonant and starting the vowel, and that delay is visible on a spectrogram in a way it is not audible to an untrained ear. It replaced judging your own production by ear while the distinction is the thing you are missing, which cannot work. Free and open source, and five seconds of looking settles what a month of listening has not.
Minimal pair drills for the three series are the other half and they have to be built rather than bought, because most learner material treats the contrast as a footnote. Take twenty pairs from a dictionary, record a native speaker or find the audio, and drill them as a listening exercise with the answer hidden. It replaced practising the sounds aloud, which trains a production you cannot yet check.
The mistake: learning the rules as exceptions
The sound changes are usually introduced one at a time, scattered through a course as things to notice about particular words, so a learner ends up with a list of oddly pronounced items rather than a system. That produces someone who knows that one specific word is not said the way it is written and is surprised by the next fifty.
Take them as a set instead, in one sitting, early. They are rules about boundaries rather than facts about words, they apply to everything, and knowing them turns every written word into a pronounceable one. It is the single most concentrated hour in Korean, it belongs in the first month, and almost every course puts it later because it looks advanced. It is not advanced. It is what turns the alphabet into reading.
What the sounds are attached to
The rules operate at the boundaries between blocks, so knowing where a block ends is a precondition, which is fourteen consonants, ten vowels, assembled into blocks. The boundaries most affected are the ones where an ending meets a stem, so the sound changes and the conjugation are the same problem seen twice, in particles, and endings that change with who you address. The words being reshaped come from two vocabularies with different registers, in native Korean, Sino Korean, and when each is used. And the last syllable of every sentence, where the ending and its sound changes land, is also where the speech level is decided, in formal, polite and casual, and the cost of the wrong one.