Every language contains a handful of sounds that some other language's speakers simply cannot produce at first. The reason is not anatomical. Infants can discriminate a very wide range of speech sounds, and that ability narrows during the first year of life as the ear specialises in the contrasts that matter in the language around it.
So the difficulty is perceptual before it is muscular. You are not failing to move your tongue correctly; you are failing to hear that anything needs to be different. Here is what that means concretely, language by language.
Czech: ř
A trilled r produced with friction at the same time — rolled and buzzed simultaneously. It appears in the composer Dvořák and in řeka, river.
It is the last sound Czech children acquire, which is a fair indication of the difficulty. Most adult learners settle for an approximation somewhere between a rolled r and the zh of "measure", and that approximation is generally understood. This is one to accept imperfection on.
Zulu: the clicks c, q and x
Three ordinary consonants that appear throughout everyday vocabulary, not emphasis or interjections. English speakers can already make all three — they just do not think of them as speech.
| Letter | Type | Familiar equivalent |
|---|---|---|
| c | dental click | the "tsk" of mild disapproval |
| q | alveolar click | a sharp pop, tongue against the palate |
| x | lateral click | the sound used to urge on a horse |
The genuine difficulty is not the click itself but producing it inside a word, at speed, without stopping. That comes with drilling rather than insight.
Arabic: ع and ح
Ayn and ḥā are made in the pharynx, further back in the throat than English uses for anything. They are the sounds most responsible for a foreign accent in Arabic, and unlike the Czech ř they carry meaning — substituting a glottal stop or an h changes the word.
These reward the approach of listening for weeks before attempting production. Learners who try to produce them from a written description usually land somewhere unrelated.
Vietnamese: the ngã tone
Vietnamese has six tones and four of them map onto pitch movements English already uses for intonation. The ngã tone does not: it rises like the sắc tone but breaks in the middle with creaky voice, a catch in the throat.
English uses creaky voice — plenty of speakers do it at the end of sentences — but never to distinguish one word from another. That is the barrier. The nặng tone, short and closed with a glottal stop, is difficult for the same reason.
Danish: stød
A creaky catch in the voice, close to a glottal stop, appearing in certain words and syllables. It is not written, and it separates words: hun (she) and hund (dog) differ largely by its presence.
Because it is unmarked in spelling, you cannot learn where it goes from the page. It has to be acquired per word from listening, which makes it one of the more demanding features on this list.
Russian: ы
A high back unrounded vowel with no English equivalent. Learners typically substitute the "i" of machine, which is audibly wrong to a Russian ear.
A technique that works for many people: say "ee" and then pull the tongue backwards without rounding the lips. It should feel slightly uncomfortable. Russian also has palatalisation — the soft sign marking a y-quality on the preceding consonant — which distinguishes мать (mother) from мат.
Hindi: aspiration and retroflexion, twice over
Hindi makes two distinctions English has physically but never uses meaningfully.
Aspiration. English aspirates the p of "pin" and not the p of "spin", and no English speaker notices. Hindi uses that difference to separate words, and the full consonant chart has a plain and an aspirated column throughout.
Retroflexion. Retroflex consonants curl the tongue back toward the roof of the mouth; dental ones put it against the teeth. English t and d sit between the two and match neither. दाल (dāl, lentils) and डाल (ḍāl, branch) differ only in this, and using an English d for both is the most common thing that makes a learner hard to follow.
Italian: doubled consonants
Included because it looks trivial and is not. Italian holds a doubled consonant noticeably longer, and the length distinguishes words: capello is hair, cappello is a hat; nono is ninth, nonno is a grandfather.
English speakers systematically under-produce these, because English consonant length carries no meaning. The fix is to lengthen past the point where it feels exaggerated — that usually lands about right.
The pattern across all of these. The hard sounds are rarely the ones absent from your language. They are the ones present but meaningless in it — aspiration, creaky voice, consonant length, retroflexion. Your ear has spent a lifetime filing them as accent, and the work is not learning a new movement but promoting a difference you already make to one you attend to.
What to actually do
- Listen before producing. Days of input before attempting a difficult sound produces better results than immediate imitation, because you cannot aim at a target you have not heard.
- Drill minimal pairs, not single sounds. Alternating two words that differ by one feature trains the contrast. Repeating one sound in isolation trains very little.
- Record yourself. Most learners believe they are producing a distinction they are not. Hearing your own attempt beside a native version is unglamorous and effective.
- Triage. Sounds that change meaning — tones, Italian length, Hindi retroflexion, Arabic pharyngeals — are worth real effort. Sounds that only mark you as foreign, like Czech ř, are worth an approximation and moving on.
Check a word before you say it
45 offline dictionary pairs with entries that preserve every diacritic and tone mark, so you know what you are aiming at. Free, and working without a connection.
Open the dictionary appFrequently asked questions
Why can I not hear the difference between two sounds in a foreign language?
Because your ear was tuned in infancy to the distinctions your own language uses and to ignore the ones it does not. English does not use aspiration or retroflexion to separate words, so English speakers hear those differences as accent rather than meaning. The ability to perceive them can be trained, but it has to be trained deliberately.
Should I fix my pronunciation early or later?
Early, particularly for perception. Pronunciation habits harden with use, and a mistake repeated for a year is much harder to correct than one caught in week three. You do not need to sound native, but you do need to hear the contrasts, because you cannot reliably produce a distinction you cannot detect.
Does an accent actually matter?
Much less than learners fear, with one exception: any sound difference that distinguishes words. An accent that makes you sound foreign is fine. A tone error that turns horse into dog, or a length error that turns hair into hat, is a different problem — that is not an accent, it is a different word.