Where an English speaker has one t, Hindi has four, and they separate real words. Two of them differ in where the tongue goes: dental, with the tongue against the back of the teeth, and retroflex, with the tip curled back to the roof of the mouth. Each of those has a partner released with a puff of air. The English t sits between the two places and carries an unpredictable amount of air, so it is heard as none of the four and as a foreign accent immediately.
Both dimensions have to be trained by ear before they can be produced, because an English ear arrives with one category where Hindi has four and cannot copy a distinction it is not registering.
Six distinctions English does not make
- ka and kha are different words. Aspiration is contrastive: a puff of air after the consonant changes the meaning, and English speakers produce it randomly without noticing.
- ta and Ta are different sounds. One is dental, with the tongue on the teeth, and one is retroflex, with the tongue curled back. English sits between the two and matches neither, which is why an English accent in Hindi is instantly placeable.
- the final a is not pronounced in modern Hindi, so the word written as naama is said naam. This is why transliterations disagree so often.
- va is between v and w, a single sound where English has two, so the same word may be written with either letter in English.
- there is no th sound as in English thin or this, and the letters transliterated th are aspirated t rather than anything like it.
- nasal vowels are marked with a dot or a crescent above the line, and they are contrastive: nasalising or not changes the word.
Two axes, and English uses neither contrastively
The place axis is the easier one to explain and the harder one to hear. English t is made at the ridge behind the teeth, between the two Hindi positions, so a learner's t is a third sound that lands roughly in the middle. Producing the dental one means moving the tongue forward until it touches the teeth. Producing the retroflex one means curling the tip back and up, which feels absurd at first and becomes ordinary within a week.
The aspiration axis is the one English speakers assume they already have. English aspirates its p, t and k at the start of a word and not elsewhere, so an English speaker produces both versions without knowing it, and produces the wrong one wherever the Hindi word disagrees with the English habit. The test is a strip of paper held in front of the mouth: the aspirated sound moves it and the unaspirated one does not, and doing that ten times is worth more than any description.
The voiced aspirates are the fourth column and the one with no English parallel at all. A voiced consonant with a puff of air after it is not something English does, and it takes deliberate work rather than adjustment.
The learner who fixes the place and forgets the breath
Here is the stall. Retroflex consonants are the famous half, they are what everybody warns about, and they are visibly a technique, so learners work on them, get them, and stop. The aspiration axis is quieter, it does not feel like a sound to be learned, and it is left to whatever the English habit produces.
What follows is a speaker whose place of articulation is right and whose breath is wrong on half the words, which is still two of the four consonants confused. Because context repairs almost all of it, nothing corrects it, and the error becomes permanent. The exit is to test the axis you are not thinking about, which is what the paper strip is for, and to drill minimal pairs by listening rather than by speaking, since a distinction the ear has not split cannot be produced deliberately. That ordering is the general rule for a sound your first language does not have, argued out in the stalls every learner hits, and the way through each.
Which errors cost a word
What changes the meaning, and what only marks the accent
Both of the first two rows separate real words, which is why they are worth a month, and the last rows are worth attention only after those are done.
Does getting this wrong change the word
- YesDental against retroflexSeparates real words, and there are many pairsYes
- YesAspirated against notSeparates real words too, on every one of the four placesYes
- YesThe dropped built-in vowelAdds a syllable, and marks you in every sentenceYes
- NoAn English r instead of the tapped oneAudible, and it costs no wordNo
- NoNasal vowels approximatedUnderstood in almost every case, and worth fixing laterNo
- Not checkedHow much any of this costs in a film soundtrackNobody has measured it, and film speech reduces heavilyNot checked
The third row is the surprise. It is not a consonant problem at all, it is a rule the script does not write down, and it marks a speaker more consistently than either of the two axes above it.
Two ways to get told
A strip of paper held in front of the mouth is the aspiration test and it costs nothing. Say the pair, watch whether the paper moves, and the answer is unambiguous in a way that listening to yourself is not. It replaced asking a native speaker whether that sounded right, which returns encouragement, because they understood you and are not going to stop the conversation over a consonant.
Hindi dictation on a phone is the other judge. Say a word and read what appears in Devanagari: a retroflex where a dental belonged produces a different word on the screen rather than a polite nod. It replaced guessing from your own ear, which hears what you intended. Free and already installed, and it should be read the same way as everywhere else: a wrong transcription is a real finding and a right one is weak evidence, since recognisers in 2026 are forgiving of accented speech.
The mistake: pronouncing the vowel the script wrote
The built-in short a in every consonant is dropped in predictable places, most reliably at the end of a word, and the rule is written nowhere. A learner who has not met it adds a syllable to a great many words and to the middle of some, which is understood every time and marks them in every sentence.
Learn the rule with the alphabet rather than after it. It takes an evening, it applies everywhere, and it is the difference between reading the script correctly and pronouncing the language, which are two different achievements in Hindi and are usually taught as though they were one. The rest of the script's honesty is described in forty six letters hanging from a line, and how they join.
What the sounds are attached to
The four-way distinction is carried in the writing by four separate letters arranged in a grid, so the script is telling you which one is meant and the spelling is not the difficulty. The words carrying these sounds come from three vocabularies, and the Persian layer brings consonants that the Sanskrit layer does not, in the Sanskrit half, the Persian half, and English on top. The endings where the dropped vowel matters most are the ones carrying gender and agreement, in gender on every noun, and postpositions where English puts prepositions. And hearing all four in fast speech, where film and news differ, is film Hindi, news Hindi, and the gap between them.