A recent analysis on Habr made a bold claim: the vowel chart of the International Phonetic Alphabet—that same trapezoid from the textbook—is wrong, and it teaches an incorrect century-old theory. The accusation is directed at a specific person: Daniel Jones, head of the phonetics department at University College London, published a scheme of "cardinal vowels" in 1917, and a few years later it turned out that X-ray images of his own articulation didn't match this scheme.
The fact itself is a one-paragraph sensation, and there's no point repeating it verbatim. What's more interesting is what happens after a canon gets debunked. A hundred years pass, and the trapezoid still hangs in every textbook and on the official letterhead of the association. So the question isn't whether Jones was wrong. The question is what replaced the measurement grid and why the replacement didn't win.
And one more detail that made me tackle this topic at all. Jones didn't just draw his vowels, he recorded them—in that same 1917, on a gramophone record. The scheme was an appendix to the sound, not the other way around. So the canon being disputed exists in two forms: as geometry on paper and as a few hundred milliseconds of sound that you can listen to today if you want.
I'll start with what's actually depicted on this scheme. The horizontal axis is frontness (front or back), the vertical is height (close or open), and the points at the corners are connected into a trapezoid. The axis names come from anatomy: height is tongue height, frontness is how far forward the tongue is advanced. That's exactly how it's explained to students.
The problem is the axes have no numerical scale. Not in hertz, not in millimeters from the palate—none whatsoever. Jones defined eight reference points and proposed considering the intervals between them equal by ear. More than eighty years later, the International Phonetic Association itself acknowledged the delicacy of the construction in the "Handbook of the IPA": the vowel quadrilateral "should be regarded as an abstraction and not a direct mapping of tongue position." The word "abstraction" here isn't from modesty—it's from the minutes.
Separately amusing is that the first real acoustic vowel chart in history appeared not in a linguistics institute but at Bell Telephone Laboratories. In 1952, Gordon Peterson and Harold Barney recorded 76 speakers—33 men, 28 women, 15 children—ran the recordings through a spectrograph, and published formant frequency tables. For [i] the first formant in men landed around 270 Hz, in women 310, in children 370. The second formant of that same [i]—2290, 2790, and 3200 Hz respectively.
That's where the first problem emerges. Jones calibrated the map to one person—himself. Formants depend on vocal tract length, and tract length in an adult man versus a child differs by about one and a half times. A point on the map meaning "vowel" actually meant "vowel pronounced by a specific London professor with a specific skull geometry."
Next—the most famous part of the story, and here I'll need to correct the retelling slightly. In the Habr account, per the instructions, "three vowels out of eight" were pronounced, the rest fitted by ear. According to primary sources it comes out differently: two were defined articulatorily—the first [i] and the fifth [ɑ]. For the first, "as far forward and up as possible, lips spread" was specified; for the fifth—"back of tongue lowered as far as possible and pulled as far back as possible, lips unrounded." The remaining six Jones constructed in auditory steps.
A difference of one unit, but it matters semantically. Jones didn't mark out corners and fit the middle to the corners—he marked two extreme points and inscribed equal-by-ear steps between them. And he stated this honestly in "An Outline of English Phonetics": "the degrees of acoustic separation between each vowel and the next are equal—or, more precisely, as equal as possible for a person with well-trained hearing."
In other words, the scheme was originally declared as an instrument of hearing, not as measurement. Students, by the way, were told this directly: in the cardinal vowel system, explains Geoff Lindsey, quality is assessed by ear, and Jones himself with his demonstrations serves as the "gold standard." The anatomical axis names—"high," "low," "front"—grew onto the scheme later, and those are what created the confusion.
Jones was not just a systematizer but an experimenter. In January 1917, physician Hugh Trevelyan George took X-ray images of his vocal tract—a chain of small lead plates was laid on the tongue so the tongue contour would be visible in contrast. This, as far as can be judged, is the very first X-ray images of articulation in the history of phonetics.
Of the eight vowels, Jones published images of four—those same edge ones. Why only four became known almost forty years later. Peter Ladefoged, still a young researcher in 1955, asked Jones directly where the rest of the photographs were. He quotes the answer verbatim: "He smiled and said, 'People would have found them too confusing.'"
With the images comes a fork where sources decisively diverge, and I'll leave it as is. According to the Utrecht collection description, four tongue images were made in January 1917 by physician Hugh Trevelyan George. Ladefoged, however, writes about images of all eight cardinal vowels, published in 1929 by Jones's colleague and namesake Stephen Jones—and it was precisely these images that showed the tongue moves between vowels not at all as the drawing looks. Whether these two series coincide (four images from 1917 versus eight from 1929) or whether these are different studies, the available texts don't clarify. Someone among the sources is mistaken, but who exactly—I can't say.
The first to voice this publicly, however, wasn't a linguist. In 1928, American speech researcher George Oscar Russell, who by then had studied about three thousand X-ray images of more than four hundred speakers—the largest articulatory study of its time—formulated the diagnosis in one phrase: "Phoneticians think in terms of acoustic fact and express the idea in terms of physiological fancy."
The phrase is good, and it should be hung over the entrance to any training course. But notice the verb: phoneticians were thinking in acoustics after all. And that's the key to the whole story.
Back to 1917. Simultaneously with the scheme, Jones released a gramophone recording of the eight cardinal vowels in long and short variants—the record came out on His Master's Voice. He made recordings twice more, for Linguaphone, in 1943 and 1956. The surviving 1956 set—two 78 rpm discs, four sides marked Eng.252–Eng.255, with a booklet for which Jones wrote a preface.
Here's the place that flips the whole moral of the plot. In that very booklet Jones writes: "The cardinal vowels can only be learnt from a teacher who can pronounce them, or from a gramophone record, or from a magnetic tape." The canon, according to the author's idea, is sound, not a drawing. The record was the real specification; the trapezoid served as a cover and cheat sheet for remembering the order.
The nicest part of this: the recordings haven't disappeared. They're digitized and sit in the open audio fund of Utrecht University—including that very 1956 when Jones was 75.
Here, though, I'll leave a "rough edge" visible that I couldn't reconcile. The Utrecht collection description says the booklet is dated March 1956, and the recordings themselves, judging by Jones's biography, were probably made between September and December of that same year. Preface before recording—a trifle, but it either means dates in the bibliography wander, or the booklet was written in advance for a ready concept. Which variant is correct isn't visible from available materials, and I won't invent.
Now—about what turned out to be "correct" instead of the scheme. Formants are resonant frequencies of the vocal tract, not peaks in the spectrum of a specific uttered sound, and this distinction matters more than it seems: a formant can "lie" between two harmonics and not be a visible peak on the graph at all. Vowels are distinguished mainly by the first and second formants, F1 and F2, with F1 roughly corresponding to "height" and F2 to "frontness" on Jones's chart.
And here begins the answer to why the scheme hasn't surrendered for a hundred years. If you plot F1 against F2 for the eight cardinal vowels—and this was done back in 1952, in the work of Delattre, Liberman, Cooper, and Gerstman—the resulting figure is shaped like Jones's trapezoid. Same slope, same elongation, same corners. Jones constructed the scheme by ear, and hearing evaluates precisely formants, so his "incorrect" drawing turned out to be a rough portrait of the real acoustic space.
Hence the moral I want to formulate more precisely than "the chart lies." The chart doesn't lie—it calls acoustic facts by anatomical words. The furniture is in place, the sign over the door is wrong. The trapezoid resembles acoustic space not by accident but because its author took measurements with an organ that measures precisely acoustics.
This also explains the strange asymmetry of the map: why exactly three vowels out of eight turned out to be rounded in the cardinal series, not four. Rounding lowers both formants, meaning it shifts the sound to a different point in space. For back vowels this carries the sound outward, to the periphery, and for front vowels—conversely, pulls it toward the center. That's precisely why rounded front and unrounded back vowels occur noticeably less often in the world's languages than others: they're all bunched in the middle, where they're hard to distinguish by ear. In Jones's scheme this looked arbitrary; in formant coordinates—a pattern.
Acoustics, however, also didn't produce a ready universal map. Here worked a third trap that debunking texts don't write about at all. Formants depend on the person: in women they're on average 15–25% higher than in men, in children even higher. To compare one speaker's vowel with another speaker's, the data must be normalized, and no canonical normalization method exists. In 2003, Peter Adank compared twelve different algorithms and got a contradictory result: methods best erasing sex and age incidentally ate away regional variation, meaning precisely what linguists want to preserve.
Professional criticism of the scheme was conducted from inside the profession and with academic courtesies. In 2001, Per Lindblad of Lund University published in his department's working papers an analysis where he collected complaints as a list. A point in the quadrilateral implies a narrow, almost instrumental definition of vowel quality, and most phoneticians don't possess such precision—meaning point notation misleads the reader. X-rays contradict the articulatory reading, especially for back vowels of non-close series, and the terms "close," "open," "front," "back" confuse students, hinting at anatomy where it's about sound.
Not even the question of what the scheme is by nature is resolved. Ladefoged insisted: an auditory instrument. Abercrombie and Catford objected: bodily-muscular, and no X-ray images are grounds here—the speaker orients by their own sensation of the tongue in the mouth. Their main counterargument sounds like mockery of pure acoustics: every secondary vowel stands on the map at the same point as its paired cardinal, though the auditory difference between them is enormous.
There's also a comparison I like best of all: Lindblad quotes Boschoff, who likened the current state of phonetics to measuring temperature before the invention of the thermometer. Sensations exist, scale doesn't. Lindblad's conclusion—abandon point notation and replace points with areas, or better yet switch to an acoustic framework based on F1, F2, and F3 ratios.
Next—a plot about how academia reacts to its own exposés. In 1988–1993 the association did conduct a scheme reform. The reform history was reconstructed in 2011 by Hartmut Pfitzinger of Kiel, and the numbers speak for themselves: a questionnaire was sent to 38 phoneticians, 24 replied on time. The question "whether to abandon the trapezoidal shape" failed with a score of 23 against 1. The question about merging two diagrams into one passed by ten votes. By separate decision they added four central vowels—as Pfitzinger writes, "with no purpose other than symmetrically filling a large empty space." He formulates his conclusion without diplomacy: recommends returning to Jones's system, and calls both reforms a regression.
Hence the answer to the original question. A replacement exists in several variants—Lindsey's acoustic grid with renormalized reference points, Boschoff's framework, three-dimensional spaces with F3 on a separate axis—and none became common language. The reason is prosaic: the trapezoid, for all its theoretical insolvency, remains the only scheme you can draw on a board in ten seconds and dictate by voice. Lindsey's proposal for standardized reference sounds, synthesized programmatically—modeled on how standard color coordinates are specified for signal lights—looks reasonable but requires everyone to switch to one synthesizer and one normalization table. Phonetics agreed on a scheme by ear in one winter of 1917 and hasn't agreed on anything since.
Every time I read about replacing the scheme, I remember the association already voted once—and got a trapezoid with four new points, which the author of the historical analysis proposes to cancel. It all ends, as usual, more modestly than it began. The IPA vowel chart is a map legend, not a map: an agreement about what symbols to use for what's already been heard by ear. Demanding precision from it is the same as demanding a compass show kilometers. It got luckier than its critics: it was drawn by acoustics and labeled by anatomy, and in a hundred years nobody bothered to change the sign.