A heteronym is a word that shares its spelling with another word but is pronounced differently and means something else. Lead the metal and lead the verb. Tear the drop and tear the rip. English keeps hundreds of them, and they are the one group of words where reading fluently teaches you nothing at all about how they sound.
What a heteronym actually is
Three terms get tangled here, and separating them takes one sentence each.
Homophones sound alike and are spelled differently. There, their, they’re. They cause writing mistakes.
Homographs are spelled alike. That is the whole requirement, so the category includes words that sound the same and words that do not.
Heteronyms are the homographs that do not sound the same. Same letters, different sound, different meaning. Bass the fish is /bæs/ and bass the low notes is /beɪs/, and no amount of staring at the four letters will tell you which one is in front of you.
The useful way to hold onto the difference is by symptom. A homophone catches you when you write. A heteronym catches you when you speak, which is a much quieter kind of catching, and that is the real subject of this article.
Why reading fluently hides them
Here is the part I find genuinely interesting. When you read silently, you are not skipping the sound. You assign one from memory, quickly and below the level of attention, and then you carry on. If the sound you assigned is wrong, nothing in the process objects.
With most English words that error eventually collides with reality. You hear the word in a podcast, notice it does not match the version in your head, and quietly correct yourself. The collision is what fixes it.
Heteronyms dodge that collision, because the wrong sound is not a non-word. It is a real English word, correctly pronounced, just not the one on the page. Read wound as /waʊnd/ in a medical article and your inner ear hears a perfectly good English word. Nothing sounds broken. There is no moment of friction to notice, so the mistake can sit there for a decade.
This is also why the problem is invisible to everyone else. A teacher can catch a spelling error by reading your work. Nobody can catch a silent mispronunciation, because it never leaves your head until the day you say it out loud in a meeting.
The largest group is a stress shift
The single biggest family of heteronyms is not exotic at all. English marks a large set of noun and verb pairs by moving the stress instead of changing the letters. Documented lists of the pattern run to at least 170 pairs, which makes it far and away the most common way an English word ends up with two pronunciations.
| Word | Noun or adjective | Verb |
|---|---|---|
| record | /ˈrekərd/ | /rɪˈkɔːrd/ |
| present | /ˈprezənt/ | /prɪˈzent/ |
| object | /ˈɑːbdʒɪkt/ | /əbˈdʒekt/ |
| subject | /ˈsʌbdʒɪkt/ | /səbˈdʒekt/ |
| contract | /ˈkɑːntrækt/ | /kənˈtrækt/ |
| conflict | /ˈkɑːnflɪkt/ | /kənˈflɪkt/ |
| increase | /ˈɪnkriːs/ | /ɪnˈkriːs/ |
| permit | /ˈpɜːrmɪt/ | /pərˈmɪt/ |
| rebel | /ˈrebəl/ | /rɪˈbel/ |
| desert | /ˈdezərt/ | /dɪˈzɜːrt/ |
Look at the second column against the third and you will notice the stress mark is underselling what happens. The syllable that loses the stress does not just get quieter. Its vowel collapses towards a neutral sound, so in reCORD the first vowel is barely there, while in REcord it is a full clear /e/. Two things move at once, and the vowel is the one listeners actually key on.
Desert is worth a second look because it changes the most. The noun is /ˈdezərt/ and the verb is /dɪˈzɜːrt/, which puts it about as far from its twin as a pair like this can get while keeping the same six letters.
The rule has a large group of exceptions
This is the part that gets left out of most explanations, and leaving it out actively creates errors. Once someone learns that nouns take first-syllable stress, they start applying it to two-syllable words that never shift, and produce things like REport and REsult.
Plenty of noun and verb pairs are spelled the same, sound the same, and keep their stress on the second syllable in both roles. Report, result, respect, reply, review, control, concern, reward, request. A report and to report are identical. There is nothing to learn about these except that they are not part of the pattern.
So the rule is not “nouns are stressed at the front.” It is closer to “a specific set of about 170 pairs marks the noun with front stress, and you have to know which words are in the set.” That sounds worse, but it is the honest version and it stops you from over-applying it.
A handful of words sit in the middle, where usage is genuinely split. Address as a noun is /ˈædres/ for most American speakers and /əˈdres/ for most British ones. Research as a noun goes both ways in the United States. When you hear a native speaker use one of these differently from how you learned it, nobody is wrong.
The -ate family is the same trap in one syllable
There is a second group that is smaller, tighter, and in my experience the one advanced learners get wrong most often. A set of words ending in -ate has an adjective or noun form ending in a reduced /ət/ and a verb form ending in a full /eɪt/.
| Word | Adjective or noun | Verb |
|---|---|---|
| separate | /ˈseprət/ | /ˈsepəreɪt/ |
| deliberate | /dɪˈlɪbərət/ | /dɪˈlɪbəreɪt/ |
| moderate | /ˈmɑːdərət/ | /ˈmɑːdəreɪt/ |
| estimate | /ˈestɪmət/ | /ˈestɪmeɪt/ |
| graduate | /ˈɡrædʒuət/ | /ˈɡrædʒueɪt/ |
| advocate | /ˈædvəkət/ | /ˈædvəkeɪt/ |
| duplicate | /ˈduːplɪkət/ | /ˈduːplɪkeɪt/ |
“A separate room” and “separate the files” are not the same word twice. The first ends in a small unstressed vowel, close to seprut, and the second ends in a full ate that rhymes with late. Say a deliberate insult with the verb ending and it lands as something you are about to do rather than something you did on purpose, which is a small error that changes the sentence.
The vowel and voicing switches
The rest of the heteronyms do not follow one rule. They are individual accidents of English spelling history, and there is no shortcut for them beyond knowing which ones exist.
- lead — /liːd/ to guide, /led/ the metal
- read — /riːd/ present tense, /red/ past tense, which is a heteronym hiding inside a single verb
- wind — /wɪnd/ moving air, /waɪnd/ to coil
- wound — /wuːnd/ an injury, /waʊnd/ the past tense of wind
- bow — /baʊ/ to bend at the waist, /boʊ/ the weapon or the ribbon
- sow — /soʊ/ to plant seed, /saʊ/ a female pig
- row — /roʊ/ a line, or to move a boat with oars, /raʊ/ a noisy argument
- live — /lɪv/ the verb, /laɪv/ the adjective
- tear — /tɪr/ a drop from the eye, /ter/ to rip
- minute — /ˈmɪnɪt/ sixty seconds, /maɪˈnuːt/ extremely small
- invalid — /ˈɪnvəlɪd/ a person made weak by illness, /ɪnˈvælɪd/ not valid
- entrance — /ˈentrəns/ the way in, /ɪnˈtræns/ to fill with delight
A smaller set switches a consonant from voiced to voiceless rather than switching a vowel. The verb use is /juːz/ and the noun use is /juːs/. Close meaning to shut is /kloʊz/ and close meaning near is /kloʊs/. Excuse works the same way, verb /ɪkˈskjuːz/ and noun /ɪkˈskjuːs/. It is the final consonant carrying the entire distinction, and it is very easy to flatten both into one sound and never know.
When context saves you, and when it does not
Most of the time context does the work, which is exactly why silent reading feels fine. By the end of the sentence you know whether the tears are falling or the paper is ripping, and your understanding is never in danger.
The trouble is that reading aloud does not get to wait for the end of the sentence. You have to commit to a pronunciation at the moment your eye reaches the word, and sometimes the disambiguating information arrives afterwards. The classic demonstration is “The bandage was wound around the wound,” where the two identical spellings need opposite vowels and the first one comes before you have any reason to know that.
Capitalization does some quiet work too. Polish the language is /ˈpoʊlɪʃ/ and polish the verb is /ˈpɑːlɪʃ/, and the only signal is a capital letter that vanishes at the start of a sentence. August the month is /ˈɔːɡəst/ and august meaning majestic is /ɔːˈɡʌst/.
None of this matters while you are reading to yourself. All of it matters the first time you present in English.
There is a good piece of evidence that this is a hard problem rather than a personal failing. Text-to-speech systems have to solve exactly the same puzzle, and heteronym disambiguation is a long-standing difficulty in speech synthesis. A screen reader converting letters to sound has no ears and no embarrassment, so it has to decide from context alone whether wound is an injury or a past tense, and it gets these wrong often enough that the problem has its own name in the field. If a system with the whole sentence in memory still struggles, a learner reading in real time is not doing badly.
The practical version is narrower than the full list. You are unlikely to be reading aloud from a novel. You are quite likely to be reading a slide, a contract, or a set of figures, and that is where the stress pairs live: contract, increase, decrease, project, record, permit, present. The high-risk words for most people are business words, which is a small enough set to be worth an hour.
If your first language does not use stress this way
English is unusual in letting stress alone carry a difference in meaning. Many languages place prominence by rule, in a fixed position in the word or phrase, so moving it produces an odd-sounding version of the same word rather than a different word entirely.
If that describes your first language, the noun and verb pairs are hard in a specific way that is worth naming. Producing the wrong stress is the smaller half of it. The harder half is that the contrast can be genuinely difficult to hear, so REcord and reCORD arrive as the same word said twice with slightly different music, and more listening on its own does not necessarily sort that out. Anything you cannot hear reliably, you cannot correct by exposure alone.
That is an argument for making the distinction visible rather than waiting to absorb it. Reading the IPA and looking at where the stress mark sits turns a difference you might not perceive into one you can see, and seeing it first makes it much easier to start hearing it. This is the same reason working backwards from sound to spelling is such a useful skill, and it overlaps with the words most commonly mispronounced by learners.
How to learn them without memorizing a list
Heteronyms respond badly to list memorization, because a list gives you two pronunciations with no hook for choosing between them. What works is smaller.
- Learn the pair, never the single word. A heteronym you know one half of is still a word you will get wrong half the time. Put both forms in front of you at once so the contrast is the thing being learned.
- Read the stress mark, not just the letters. In IPA the mark sits before the stressed syllable. Once you look for it deliberately, /ˈrekərd/ and /rɪˈkɔːrd/ stop looking like the same word.
- Say both out loud, back to back. This is the step people skip and it is the one that does the work. Producing the contrast is what makes it audible to you later.
- Come back to it after you have forgotten. A heteronym learned once reverts to whichever pronunciation you met first. It needs to be re-met on a schedule.
Vocaby carries IPA and audio for every one of its 29,000-plus English words, so both members of a pair are one tap from being heard rather than guessed at, and the search matches any part of a headword if you only half remember the spelling. Reviews are scheduled with FSRS, which brings a word you keep getting wrong back sooner and one you have settled further out. You can also browse the word explorer directly if you would rather look things up than study them.
The thing worth taking away is smaller than the word count suggests. Fluent reading is not evidence that you know how a word sounds, and for a few hundred English words it is not even weak evidence. Those words are findable, they cluster in two large groups, and an afternoon with the stress pairs will retire most of the risk.
Learn English vocabulary with Vocaby on the App StoreFrequently asked questions
- What is a heteronym in English?
- A heteronym is a word that shares its spelling with another word but is pronounced differently and means something different. Lead the metal is said /led/ and lead the verb is said /liːd/. Heteronyms are a subset of homographs, which are words spelled alike whether or not they sound alike.
- What is the difference between a heteronym and a homophone?
- They are opposites. Homophones sound the same and are spelled differently, like there and their. Heteronyms are spelled the same and sound different, like the two versions of tear. A homophone trips you up when you write. A heteronym trips you up when you speak.
- Why do nouns and verbs like record have different stress?
- English marks a large group of noun and verb pairs by moving the stress rather than changing the letters, so the noun is REcord and the verb is reCORD. Documented lists of this pattern run to at least 170 pairs. The unstressed syllable usually reduces to a neutral vowel as well, so the two forms differ by more than the beat.
Keep reading
English Words That Changed Meaning Completely
Nice once meant foolish. Awful once meant awe-inspiring. Meat was any food at all. Here is how English words drift, the patterns the drift follows, and why a moved meaning is harder on learners than a word you have never seen.
Which Parts of Your Accent Actually Cost You, and Which Are Free
Nobody needs a native accent. But some pronunciation errors genuinely stop people understanding you, and others cost nothing at all. The useful skill is telling them apart instead of trying to fix everything.
English Words Borrowed From Other Languages, and Why They Break the Rules
English took words from French, Latin, Greek, German, Arabic and Japanese, and each source left fingerprints on the spelling. Once you can read those fingerprints, a lot of English pronunciation stops looking random.