A listener locates an English word by its beat before the vowels have finished arriving. Put the beat on the wrong syllable and they do not hear a foreign accent; they hear a word they have never met, which is why the response is a request to repeat rather than a guess.
A wrong vowel usually leaves the word findable and slightly odd. A wrong beat removes it from the listener's search entirely, and the sentence stops while they work backwards from the context to recover it.
Eight facts about where the beat falls
- stress errors cost more than sound errors. A word with every sound right and the stress wrong is often not understood at all, while the reverse usually is.
- PHOtograph, phoTOGraphy, photoGRAPHic. One family, three stress positions, and the vowels change completely each time.
- stress moves the vowels. Unstressed syllables collapse to schwa, so knowing the stress tells you most of the pronunciation.
- REcord and reCORD. Noun on the first syllable, verb on the second, and around thirty common words behave this way.
- two syllable nouns usually stress the first, and two syllable verbs usually stress the second, which is the pattern behind the rule above.
- -tion pulls stress to the syllable before it, reliably: eduCAtion, inforMAtion, decoRAtion. The same is true of -ic and -ity.
- comfortable has three syllables in ordinary speech, not four, because the middle one disappears entirely.
- vegetable has three syllables for the same reason, and so do chocolate, different and interesting.
The beat is part of the word, stored with it
English stress is not something added to a word for effect. It is a property of the word, learned with it, and as fixed as its spelling. Banana has its beat on the middle syllable in the same way it has an n in the middle, and there is no situation in which a speaker chooses otherwise.
Nothing about the spelling tells you where it goes, and English gives no reliable rule for it. There are tendencies, most of them tied to endings rather than to roots, and the honest summary is that stress has to be learned per word and per derived form. A dictionary marks it and almost nobody looks at the mark.
Where it stops: learning it in a chapter about emphasis
The stall is caused by where this normally gets taught. Stress arrives in a lesson about emphasis, alongside saying a word louder to make a point, and that framing tells a learner it is optional decoration. Everything that follows from that is wrong.
A learner who believes stress is emphasis will move it. They will beat the first syllable of a word to underline it, and produce something that is not a word at all, and then be surprised that emphasising a point made them less clear rather than more. Meanwhile the actual job, storing the beat with each new word as part of the word, never starts, because nothing suggested it was necessary.
The correction is one sentence long and it changes what you write in your notes: record the beat when you record the word, or you have recorded most of a word.
Where stress alone carries the word class
English has a set of pairs where nothing but the beat separates a noun from a verb. REcord and reCORD, PREsent and preSENT, OBject and obJECT: same letters, same sounds in the same order, two different words and two different parts of speech.
The pattern puts the noun's beat on the first syllable and the verb's on the second, and it is worth knowing. It is also not a rule, which is the part usually left out. PROmise, ANswer and VIsit are noun and verb with no shift at all, and nothing in the words predicts which behave one way and which the other.
Checking a beat in five seconds, anywhere
The speak button in Google Translate is the most available instrument here rather than the best one, and availability is the argument: it is on every phone and in any browser, it needs nothing installed, and it will say a word aloud with its standard stress in about five seconds. It replaced hoping that a coursebook recording happened to contain the word you were unsure about.
Its limit falls exactly on the pairs above. Given record on its own it will choose one of the two and cannot know which you meant, so for that set you have to feed it a sentence rather than a word. For everything else, which is almost all of the vocabulary, a single word is enough.
What it looks like when it goes wrong
The recognisable failure is a correctly derived word said with the root's beat, which produces something the listener hears as unknown rather than as mispronounced. The giveaway is the reaction: not a correction, but a pause and a request to say it again.
The quieter failure is on the pairs. Saying reCORD where the noun was meant is understood from context every time, so it never gets repaired, and it goes on marking the speaker for years.
Which syllable loses its full vowel once the beat is placed elsewhere is the twenty that English hears and most languages do not, and the way endings move the beat across a whole family of words is one root, and the nouns, verbs and adjectives around it. Both sit under being understood the first time you say it.
The open question is what happens to all this once the word is inside a sentence, because a sentence has its own beats and they do not simply add up from the words: most of the words in a spoken sentence surrender their stress entirely. That is why English stresses some words in a sentence and swallows the rest.