Papers
Where the beat came from: a short history of rhythm in language teaching
Every teacher has met the student who knows the words and still cannot be understood. The grammar is there, the vocabulary is there, and the sentence still arrives as a row of separate bricks. For about a century, teachers and phoneticians have kept returning to the same suspicion: what is missing is not knowledge, it is rhythm.
The idea that English has a beat
Mid-twentieth-century phonetics gave the idea its first shape. English was described as a stress-timed language: stressed syllables fall at roughly even intervals, and whatever sits between them gets squeezed. Spanish, by contrast, was described as syllable-timed, with syllables of more even length. The division is usually traced to Pike (1945) and Abercrombie (1967), and it was challenged almost as soon as it became popular: measurements of real speech did not find the even intervals the theory promised, and later work (Roach 1982, Dauer 1983) argued the classes are tendencies produced by syllable structure and vowel reduction, not clockwork.
That argument is still open, and it matters for teaching more than it looks. If English rhythm is carried by compression of unstressed syllables, then "what do you want to do" is not six words for the ear, it is one shape with two peaks. A student who pronounces all six words with equal weight is speaking a language nobody in the room is listening for.
Chants: putting a sentence on four beats
The classroom answer arrived in 1978, when Carolyn Graham, who taught English as a foreign language at NYU for twenty-five years, published "Jazz Chants: Rhythms of American English" with Oxford University Press. The method is simple enough to describe in one sentence: take a natural spoken phrase, keep its real stress and intonation, and lay it over a steady four-beat pulse.
Graham's complaint about textbooks was not that they were too hard but that nobody talks like that. Her recipe put the burden on authenticity first: a topic students actually care about, real language, a short vocabulary chant built from words of different lengths, then, if you want it, a grammar pattern inside the same frame - he plays, she plays, they play. The structure that followed became the standard shape of a grammar chant: one fixed rhythmic frame, one slot that changes, call and response between leader and chorus, contrast pairs that force the ear to notice a difference, and a repeat at the end that fixes the form.
A series followed from the same publisher, and chants became the backbone of a full course ("Let's Go", which Graham co-authored), with drum tracks next to the songs. The method spread because it needs nothing: no equipment, no screen, a teacher and a room full of voices.
Two notes for anyone writing their own material. The method itself is free - rhythmic speech over a pulse, question and answer, a grammar pattern in a frame. What belongs to the publisher is the specific texts, the recordings and the name, which Oxford University Press marks as a registered trademark. Write your own chants, and call them chants.
Songs, karaoke and the machines
The second line of development came from music. Songs had always been in classrooms, and from the 1970s onwards the methods built around relaxation, music and physical response made them respectable. The modern form is karaoke: text synchronised to a recording, the learner joining in.
Here the technology moved faster than the teaching. Today a learner can get song lyrics lined up by line (and in one commercial player, by syllable), can fill gaps in the text at four different densities, can read translations under every line. What almost nobody does is listen back: in the whole landscape of language-learning products built on songs, the learner's voice is essentially never assessed. Where singing is scored - in music apps and in mass-market karaoke - what is measured is pitch and hitting the beat, not words and not pronunciation.
The reason is technical and worth knowing, because it is the reason chanting remains more useful than singing for feedback. Recognising sung words is still hard: the best results reported in the MIREX 2020 evaluation were around 47 per cent word error on songs with music, against 5 to 10 per cent on clean speech. Backing music makes it worse, which is why every karaoke developer asks for headphones. Speech engines are built for speech; they suppress non-speech sound and work best with a quiet background.
So the two traditions end up with a division of labour that no marketing page states plainly. Singing is the stronger of the two for remembering material. Rhythmic speech over a beat is the one a machine can honestly judge.
What the history leaves us
A hundred years of work has produced a method that is cheap, portable, widely used and genuinely enjoyable - and a research record that is much thinner than the method's popularity suggests. That gap is the subject of the next article: what the evidence supports, what it does not, and which questions nobody has answered yet.
Sources behind this article: Pike (1945), Abercrombie (1967), Roach (1982), Dauer (1983) on speech rhythm classes; Graham, "Jazz Chants" (Oxford University Press, 1978) and "Creating Chants and Songs" (OUP, 2006); MIREX 2020 lyrics transcription results; product review of song-based learning tools, November 2026.
All articles
Free practice for every rule
Every grammar rule on the site has a trainer behind it. It is free and needs no sign-up.