150th Anniversary of the Bell Vowel Model 4
September 2017 saw the 150th anniversary of Alexander Melville Bell’s vowel model. Daniel Jones’ cardinal vowel system was a modification of Bell’s model, especially reducing Bell’s three low vowels to two. Was his open front really open front?
This topic is embarrassing – Daniel Jones’ cardinal vowel 4, [a], appears to be open central and not the open front he had intended, excluding an area for vowels that are potentially more front than cardinal 4. An example is the new Southern British English (SBE) open TRAP vowel with higher F2 than cardinal 4, indicating it lies beyond the cardinal diagram (Figures 1-3). This sound change has been noticed for Received Pronunciation (RP) since about 1960, for example Bauer (1985), Deterding (1997) or J. C. Wells (1962, 1982). However, there is evidence it was already occurring earlier in the 20th century RP (Fabricius 2007) and before 1900 in regional Kentish SBE (Wood 2017). Other examples are Egyptian Arabic and West Greenlandic, both with open vowel allophones potentially more front than cardinal 4.
– Bauer, Laurie. 1985. Tracing phonetic change in the received pronunciation of British English. Journal of Phonetics 13:61-81.
– Deterding, David. 1997. The formants of monophthong vowels in Standard Southern British English Pronunciation. Journal of the International Phonetic Association 27:47-55.
– Wells, J. C. 1962. A Study of the Formants of the Pure Vowels of British English. Unpublished MA thesis, University of London (online).
– Wells, J. C. 1982. Accents of English. Cambridge University Press (3 vols).
– Wood, Sidney A. J. 2017. A spectrographic study of sound changes in nineteenth century Kent. In Tsudzuki, Masaki & Masaki Taniguchi (eds), A Festschrift for Jack Windsor Lewis on the Occasion of his 90th Birthday 215-216, Journal of the English Society of Japan 21. (online)
Figure 1. F1/F2 diagram for RP peripheral monophthongs: Sir Richard Paget (1869-1955, Eton College, Oxford University, scientist and barrister, sign language promoter, British Deaf Association), showing three different unrounded open vowels (F1>600Hz): (1) open TRAP, (2) [a]-like STRUT, (3) [ɑː]-like BATH. The formants were analysed by Paget himself (1930, see §2.1 below) and reported as F1 and F2 ranges from which this diagram was made (an example at GOOSE, top right).
Figure 2. F1/F2 diagram for RP peripheral monophthongs by Stanley Baldwin (1867-1947, Harrow School, Cambridge University, industrialist, politician), taken from a political broadcast. He has three different unrounded open vowels (F1>650Hz): (1) open TRAP, (2) [a]-like STRUT, (3) [ɑː]-like BATH).
Figure 3. F1/F2 diagram for RP peripheral monophthongs by Daniel Jones (1881-1967, Radley College and University College School, Cambridge University, professor). Taken from two gramophone records, a 1929 Linguaphone language dialogue and the commentary to the 1956 cardinal vowel recording. Both show three different unrounded open vowels (F1>650Hz): (1) open TRAP, (2) [a]-like STRUT, (3) [ɑː]-like BATH). The 1929 and 1956 renderings were similar, except for STRUT, BATH and LOT where the regressions included higher F1 in 1956.
This overview is concerned with unrounded open vowels, but a new formal definition of open (or low) is needed now that the conceptual basis of the Bell vowel model is questioned (see The Bell vowel model and The double resonance theory). Wood (1979, 1982) found that the vocal tract is constricted by the tongue body at one of four locations for vowels (the hard palate, the velum, the upper pharynx and the lower pharynx). Figure 4 shows the F1 and F2 generated in a model vocal tract by varying the degree of constriction and degree of mouth opening at each of these four locations. Of relevance now is that vowels typically classed as unrounded open have F1 higher than about 600Hz (F1 is slightly lower for rounded open back [ɒ] due to the lip rounding, see the LOT examples in Figures 1-3). The use of [æ] in Figure 4 also needs explaining. The IPA [æ] has always been located where Jones needed it for the closer 19th century RP TRAP, halfway between [ɛ] and [a]. There has never been a dedicated character for the new open TRAP of 20th cenrtury Home Counties SBE and recent RP, some authors (including me) substituting [æ], others [a]. I refer to the IPA [æ] as closer [æ] and the substitution as open [æ].
– Wood, Sidney A. J. 1979. A radiographic analysis of constriction locations for vowels. Journal of Phonetics 7:25−43.
– Wood, Sidney A. J. 1982. X-Ray and Model Studies of Vowel Articulation. Working Papers 23, Department of Linguistics, Lund University.
Figure 4. Vowel F1 and F2 obtained by varying the degree of constriction (Amincm2) and the degree of mouth opening (A/lcm) at the four constriction locations (palatal, velar, upper pharyngeal, lower pharyngeal) in a simple model vocal tract, based on the Stevens & House (1955) nomograms. A selection of vowel timbres has been added, spaced approximately for this model ([æ] is open, not the closer IPA [æ]). In this example, the open vowels (low pharyngeal constriction) have F1>600Hz. In real life, this boundary varies individually depending on vocal tract size and speaking style, usually around 600~700Hz. The West Greenlandic informant (Figure 9) raised this boundary by about 100Hz in continuous speech. Adapted from Wood (1979).
Figure 3 records Daniel Jones’ TRAP as open, with F1 higher than about 650Hz and extending to about 850Hz. His F2 for TRAP in Figure 3 is as high as 1600-1800 Hz, suggesting it represents his high F2 range for open vowels while cardinal 4 is in his mid F2 range (Figure 5, 1376-1685Hz, midpoint 1580Hz,). This is the main reason for questioning his own belief that cardinal 4 would be the front-most open vowel. However, the three examples in Figures 1-3 are controversial since they were born in the 19th century between the 1860s and 1880s, long before anyone spoke of a TRAP sound change in RP. Jones never mentioned it in (1932) or in later revised editions up to the 1960s, describing only the earlier closer timbre. To decide this issue while avoiding controversial data, seven other recordings of RP will be examined from prominent people born in the 20th century between 1900 and 1930. Fabricius’ oldest subject was born in 1900 and had the new open TRAP, while others born in the 1920s were still acquiring the older closer TRAP, suggesting this sound change was already in progress then. Both old and new TRAP pronunciations can be expected from RP speakers born in the 1920s.
– Jones, Daniel. 1932. An Outline of English Phonetics. Leipzig: Teubner (3rd edition).
Figure 5. The spectral zones of primary cardinal vowels (P1-P8) by Daniel Jones (1956 version), sampled once during each instance. The elongated ellipses enclosing the plots suggest underlying regressions that can be explained by reference to Figure 4.
This is not an occasion for doubting Jones’ integrity. However, this issue clearly focuses on his handling of vowel theory and has to be resolved. Various aspects of his cardinal vowel scheme have been criticized in the past (for example Butcher 1982, Ladefoged 1967, Lindblad 2001) and today, a century after Jones’ published X-ray photographs of cardinal vowels in (1917), we should be able to view his work in perspective. He was obviously having difficulties after 1917, attempting to reconcile the evidence of his X-ray images with his expected tongue heights (Collins & Mees 1999:§7.6). A. M. Bell’s vowel model (1867:15-16), with its innovative but hypothetical continuum of tongue positions between back and front, was accepted universally by 1900 when Jones began studying phonetics. As part of the cardinal vowel scheme, Jones adapted Bell’s model to Passy’s French vowel diagram. Some of his modifications were little more than cosmetic like renaming the labels high–low to close–open and mixed to central. But others had fundamental consequences, like reducing Bell’s three low vowel positions to two, or transforming Bell’s matrix (where low comprises one third of the height range, emphasizing generalized zones) into a Cartesian coordinate system (where open is at the lower edge, emphasizing precise locations). Today we know the Bell model has never been validated during its 150 year existence, while contradictory evidence from X-rayed vowel profiles was accumulating from around 1900 onwards culminating with Russell (1928). The Bell model was already compromised by 1917-32 when Jones was introducing his cardinal vowel scheme, but no solution was available until the work of Chiba & Kajiyama (1941) and of Fant (1960) some 20 to 30 years later. All this means that tongue height and backness can no longer be used in the sense Bell originally conceived for them.
– Bell, Alexander Melville. 1867. Visible Speech. London: Methuen.
– Butcher, A. 1982. Cardinal vowels and other problems. In David Crystal (ed), Linguistic Controversies. London: Arnold. 50-72.
– Chiba, Tsutomu & Masato Kajiyama. 1941. The Vowel, its Nature and Structure. Tokyo: Tokyo-Kaiseikan. Reprinted 1958, Tokyo, Phonetic Society of Japan.
– Collins, Beverley & Inger M. Mees. 1999. The Real Professor Higgins. Berlin: Mouton De Gruyter.
– Fant, C. Gunnar M. 1960. The Acoustic Theory of Speech Production. The Hague: Mouton.
– Jones, Daniel. 1917. Experimental phonetics and its utility to the linguist. Nature 100:96-98.
– Ladefoged, Peter. 1967. Three Areas of Experimental Phonetics. London: Oxford University Press.
– Lindblad, Per. 2001. Critique of the IPA vowel quadrangle. Working Papers 49:104-107. Department of Linguistics, Lund University.
– Russell, G. Oscar. 1928. The vowel, its physiological mechanism as shown by X-ray. Columbus (Ohio).
The failure of the Bell model to account for vowel articulation (applying equally to Jones’ version of it), and of the double resonance theory (that had seemed to justify the Bell model for a few decades) was noted above. Following Essner (1947) and Joos (1948), front will be interpreted as high F2, central as mid F2, and back as low F2 (with actual numerical values depending on the vocal tract length of individual speakers). Close (high) is interpreted as low F1, and open (low) as high F1.
– Essner, C. 1947. Recherches sur la structure des voyelles orales. Archives Néerlandaises de Phonétique Expérimentale 20:40-77.
– Joos, Martin. 1948. Acoustic Phonetics. Language Monograph 23, supplement to Language 24.
Jones had likened cardinal 4, [a], to “the Fr[rench] sound of a in la” spoken by “educated Parisians” (1932:§144, repeated in all later editions). Clearly, he never expected to find an open vowel more front than Parisian French /a/. No-one seems to have questioned the location of cardinal 4, unless I have missed them, although I did find a few signs that individual researchers might have addressed this problem quietly without fussing about the cause. An earlier example was Delattre et al. (1952) who put [æ] at open front with high F2 and [a] at open central with mid F2, for a perception experiment with synthetic cardinal vowels. A more recent example is Fougeron & Smith (1993) who, without comment, put their Parisian /a/ at open central and not at open front where Jones had always wanted it.
– Delattre, Pierre, Alvin M. Liberman, Franklin S. Cooper & Louis G. Gerstman. 1952. An experimental study of the acoustic determinants of vowel color. Word 8:195-209.
– Fougeron, Cecile & Caroline Smith. 1993. Illustrations of the IPA:French. Journal of the International Phonetic Association 23:73–76
The queried location of cardinal 4 is reflected in requests for a symbol for the vacant open central vowel in the IPA vowel diagram, and in the debate at the 1989 Kiel Convention (IPA 1989). These requests have always been refused. In a recent instance, Barry & Trouvain (2008) proposed explicit identification of open central, while Recasens’ (2009) solution was to substitute a phonetic character that happens to be unused, the procedure advocated by Jones. I checked four random issues of JIPA, containing some dozen Illustrations of the IPA, and all the authors were substituting [a] from cardinal open front to open central. Barry & Trouvain (2008) collected many more examples of current substitution practices by JIPA authors. These substitutions emphasize the confusion arising from uncertainty about the location of cardinal 4. Strictly speaking, with French /a/ at open front and French /ɑ/ at open back, open central would have had little meaning and was consequently rarely shown on Jones’ own diagrams. The unnamed open central position marked on the IPA vowel diagram today is a recent innovation that was not included in Ashby’s (1989) review or recommendations.
– Ashby, Michael. 1989. A note on the vowel quadrilateral. Journal of the International Phonetic Association 19:83-88.
– Barry, William J & J. Trouvain. 2008. Do we need a symbol for a central open vowel? Journal of the International Phonetic Association 38:349-357.
– IPA. 1989. Report on the 1989 Kiel convention. Journal of the International Phonetic Association 19:67-80.
– Recasens, Daniel. 2009. Response to W. J. Barry & J. Trouvain, Do we need a symbol for a central open vowel? JIPA 38 (2008), 349- 357. Journal of the International Phonetic Association 39:231-233.
Phonetics literature from the late 19th and early 20th centuries has been studied in order to capture some of the theoretical background preceding and surrounding Jones’ first presentation of some of his X-rayed cardinal vowel profiles in (1917) and the definitive version of the cardinal vowels fifteen years later in the revised Outline of English Phonetics (1932). For example, Jespersen (1889:18-19) had speculated that there should never be any low central vowels because the tongue ought to make a discrete transition between front and back for low vowels. This aspect of the debate seems to be echoed by Recasens (2009) favouring Jones’ shorter [a]-[ɑ] base for the cardinal diagram. Recognising a vowel that is more front than cardinal 4 would not have been a problem in the Bell vowel model (with three low vowels), but it was critical in Jones’ modified version (with only two open vowels). Sweet (1877:16) put the trap vowel of man at Bell’s low front without mentioning any sound change, while Jones put [æ] at the earlier trap location nearer dress, his lowered half-open, where it remains today in the IPA vowel chart.
– Jespersen, Otto. 1889. The Articulation of Speech Sounds Represented by Means of Alphabetic Symbols. Marburg: N. G. Elwert.
– Sweet, Henry. 1877. Handbook of Phonetics. London, MacMillan.
The dubious claim that tongue positions could be felt and determined from muscular sensation appeared soon after the Bell vowel model, and was reviewed in Feeling tongue positions. Perturbation theory (defining the relationship between vocal tract shape and formant frequencies) will be outlined in a forthcoming page. Finally, Jones had a shorter than average vocal tract (shown by his higher formant frequencies generally), which has to be allowed for when evaluating the formant frequencies of his regular speech and of his cardinal vowels.