The ancient throat-tongue-lip model
- The two classes of model
- Ancient India
- Ancient Rome
Before Alexander Melville Bell introduced his new vowel model in Visible Speech (1867), vowel production had been conceptualized as a tree: a tongue (or palatal) branch for [i e]-like vowels, and a lip (or labiovelar) branch for [u o]-like vowels, splitting from a throat node for [a]. Intervals along the branches marked mouth (or jaw) openings. This throat-tongue-lip model was inherited from the times of Panini in ancient India, spreading east to China and Japan and west to Arabia through Buddhism and cultural contact with Islam, before reaching Greece and Rome and the rest of Europe (Beal 1906, Staal 1972, Fleisch 1957, Gairdner 1935). Along the way, a tongue-lip (or mixed) branch was added for [y-ø]-like vowels (combining palatal tongue position with lip rounding). Bell had himself followed what was still the tradition of his time in 1849, by including the ancient model in his Principles of Speech.
1. The two classes of model
The Bell model sees the mouth as a free space defined as a coordinate system where each vowel has a unique location. Any slight shift of location in any direction creates a new vowel. Bell’s revolutionary innovations were (i) the concept of tongue height, (ii) the concept of central tongue positions between front and back, and (iii) the concept of small increments of tongue position in any direction (Fig. 1 left).
Figure 1. Comparison of the Bell model (left) and the throat-tongue-lip model (right).
In contrast, the ancient model divided vowels into pharyngeal (throat or gutteral) [a]-like vowels, palatal (tongue) [i-e]-like vowels, and lip [u-o]-like vowels (Fig. 1 right). A velar tongue position for [u] was not always stated, but this does not imply ignorance of its labio-velar character. For example, it is implicit in the 4th century AD account of Marius Victorinus, and explicit by Hellwag near the end of the 18th century.
Figure 2. The throat-tongue-lip model in its final and most complete form: branches for labiovelars, unrounded velars, rounded palatals, unrounded palatals.
Extra branches were added as required, for rounded palatals by the renaissance and plain velars during the 19th century (Bell claimed to have discovered them but they were already established).
Figure 3. Examples of 19th century throat-tongue-lip trees, from Hellwag (1781), Chladni (1809) and Du Bois-Reymond (1812).
Conceptually, the ancient model is a tree, with linear tongue or lip series branching from a throat node. This structure was expressed verbally from the beginning, then by the 18th and 19th century iconically in a variety of shapes from vertical or horizontal tree diagrams (Fig. 3) to pyramids, and cruciform or radial versions. The extra branches for rounded palatals and plain velars were known as mixed because they combined the original tongue and lip branches in different ways (Fig. 2). The intermediate location of the mixed branches in any graphic variety of the tree never expressed intermediate tongue positions between front and back.
Figure 4. John Wallis (1653) expressed the throat-tongue-lip model in matrix form: (1) gutterals, palatals and labials, vs. (ii) degrees of jaw opening.
Wallis (1653) presented his version of the ancient model (Fig. 4) in the form of a matrix, which prompted Michaelis (1881, p. 411) to welcome it as a major step forward, on the strength of its square format like Bell’s new model. Unfortunately, Michaelis was confusing iconographic form with conceptual content. Wallis’s account continued the parameters of the ancient model without modification – three places of location (gutteral, palatal and labial), and three degrees of jaw opening (explicitly jaw opening, apertura faucium). Wallis foresaw the possibility of undiscovered or future languages at each constriction location, but he never hinted there might be more tongue locations in between.
Examples of the long tradition of the ancient model are found in the 6th and 5th century BC Indian recitation manuals (Whitney 1862 and 1871, Regnier 1856, Ghosh 1938), the Roman grammarians Terentianus Maurus and Marius Victorinus (Keil 1855-1880 vol. 6), Arab grammarians such as Avicenna (Ibn Sina) and Ibn Ginni (Bravmann 1934, Semaan 1963, Fleisch 1958). In post-renaissance Europe there are examples like John Hart (1569), Jacob Madsen af Aarhus (1589), John Wallis (1653) and later still Hellwag (1789), not forgetting Bell himself (1849) and Helmholtz (1863).
2. Ancient India
The phonetics literature consists of the pratisakhyas particular to each Veda and the siksas dealing with general phonetic topics (Whitney 1879, Keith 1909, Varma 1929, Allen 1953). The most comprehensive is the Paniniya siksa (Ghosh 1938) which is not usually ascribed to Panini himself (except by Ghosh who also claims it to be of very ancient date). The following chronology has been given to the treatises consulted:
6th century BC:
(RP) Rig-Veda Pratisakhya, (Regnier 1856-58)
(TP) Taittiriya Pratisakhya (Whitney 1871)
(AP) Atharva-Veda Pratisakhya (Whitney 1862)
5th century BC:
Astadhyayi (Panini’s grammar) (Böhtlink 1887, Vasu 1897)
(PS) Paniniya siksa (Ghosh 1938)
These treatises describe a as a throat vowel, i as palatal, and u as labial. The intermediate vowels [e] and [o] are the sandhi reflexes of /a+i/ and /a+u/ respectively. Regarding place of articulation and active articulator, The TP declares that “in the case of the vowels, that is the place of their production, to which approximation is made” and “that is the producing organ which makes the approximation”.
2A. Throat vowels
The TP prescribes “in the absence of special direction the tongue is thrust down forward”. Regnier defines the actual Sanskrit term kanthya as “naît dans la gorge, qui a pour lieu, pour organe, la gorge” and states that “la voyelle a et h sortent de la gorge, tandis que le k, le g etc. viennent de l’entrée de cette organe, c’est à dire de la racine de la langue” (and Allen explains that the Sanskrit term for the root of the tongue, jhivamula, refers to the dorsal part and not to what we now know as the root). All the treatises link a with h, which prompted Allen to interpret kanthya as ‘glottal’.
2B. Palatal vowels
The RP states that “la lettre e, l’ordre qui commence par c, les lettres i et ai, le y, le s, sont des palatales”. The AP adds that “of the palatals, the middle of the tongue is the producing organ”, and the commentary enumerates the relevant sounds, including e, ai and three quantities of i. The TP states that in the i-vowels, the middle of the tongue is to be approximated to the palate, also in e“. The TP specifically mentions that for e “one touches the borders of the upper back jaws with the sides of the middle tongue” recording lingual contact with the molars during palatal articulations.
2C. Labial vowels
The vowels o and u are listed together as one series with the labial consonants. The TP states that in the u-vowels there is “an approximation of the lips” and “for o the lips are more nearly approximated” than for a. The AP gives the common characteristic of all labials that “the lower lip is the producing organ” and the commentary lists the vowels o, au, and three quanties of u, along with all the labial consonants.
Aperture ranges from complete occlusion for labial stops to wide open for a. The AP states that the articulator, in the case of vowels “is open”, continuing “in the case of e and o it is very widely open and even more so in the case of a“.
3. Ancient Rome
Only two Roman authors appear to have left a surviving account of vowel articulation, Terentianus Maurus from the 2nd century AD and Marius Victorinus from the 4th century AD (Keil vol 6).
Terentianus Maurus wrote a manual in verse, De litteris, syllabis et metris Horatii, in which he describes the throat-tongue-lip model. The basic or natural vowel a involves maximum depression of the jaw and tongue: for e and i the tongue involves maximum depression of the jaw and tongue; for e and i the tongue is pressed upwards and the mandible raised; the tongue does not touch the upper teeth at all for a, whereas it is pressed against the molars for e, and all the molars and side teeth for i. New detail concerns tongue retraction. The “tragic tone of the mouth cavern” of o and the “graver quality” of u are enhanced by rounding and protruding the lips.
Marius Victorinus has a chapter De enuntione litterarum in his Ars grammatica in which he gives a similar account: “… The letter a is produced with the mouth wide open and the tongue lowered and not touching the teeth; e, which follows, is uttered with a moderate lowering of the jaw and the lips drawn in; i will have the mouth half closed and the tongue pressed lightly against the teeth; … o, like e, emits a two-fold sound according to its duration … thus the short vowel will have its lips opened a little and the tongue withdrawn; the long vowel, however, will produce its tragic sound with protruded lips, rounded mouth and the tongue hanging in the cavern of the mouth; … whenever we utter the letter u we produce it with lips protruded and drawn together …”.
4. The renaissance
The grammars of Donatus and Priscian were well known textbooks throughout the middle ages and were important agents by which the classical tradition was transmitted to later centuries, but they were not a source of phonetic knowledge. Donatus mentions neither Terentianus Maurus or his contemporary Marius Victorinus. There was only one known copy of Terentianus Maurus’s treatise during the middle ages, in the Lombardian monastery of Bobbio (Sandys 1906).
By the 16th century, vowel articulation was again being described in terms of pharyngeal, palatal and labial constrictions or gestures, for example in John Hart’s Orthographie (1569, Danielsson 1955). At least one post-renaissance grammarian, Bishop Jacob Madsen af Aarhus (1586), frequently quotes from Terentianus Maurus’s treatise, indicating that it was in circulation again.
Finally, the ancient model found its way into western linguistics by new routes, via the Arab world in the middle ages, via trading and colonial contacts with the orient (Fleisch 1957, Gairdner 1935, Staal 1972). In one form or another, the throat-tongue-lip model survived for two to three thousand years until the end of the 19th century, when it was rejected by Bell in the name of science, and discarded by his contemporaries. I suggest instead that it was based on verifiable visual and tactile observations of articulator gestures and constriction locations that also turn out to be spectrally relevant in the light of an adequate acoustic theory. It is Bell’s innovations that have not been empirically validated and that have no spectral significance outside the primitive single cavity theory that inspired them, but that’s a separate story.
This page summarizes part of X-Ray and Model Studies of Vowel Articulation, The history of the classical vowel articulation model, and The throat-tongue-lip model of vowel articulation.