RP SBE vowel formants

Introduction

This page introduces a spectrographic study of the monophthong vowels of five speakers of RP (the same five speakers who were introduced in Southern British English). Together, they cover 100 years (birth years from 1874 to 1961), offering snapshots of some of the changes reported in RP during that period.

Additionally, there is similar data reported from the Queen’s Christmas Broadcasts by Jonathan Harrington et al. (2000, Monophthongal vowel changes in Received Pronunciation: an acoustic analysis of the Queen’s Christmas broadcasts, Journal of the International Phonetic Association 30: 63-78).

Finally, there is similar data published by David Deterding (1997, The formants of monophthong vowels in Standard Southern British English Pronunciation, Journal of the International Phonetic association 27, 47-55). Deterding doesn’t define or cite a source for Standard Southern British English, except to say it was “similar to RP”, and assumes, following Peter Roach et al. (1993, MARSEC: a machine-readable spoken English corpus, Journal of the International Phonetic Association 23, 47-54), “that the accent of all the speakers is RP or close to it”. In turn, I understand this to mean these accents are either truly RP, original or successfully adopted, or some other accent(s), partially modified towards RP. That might also include home counties SBE, perhaps modified towards RP, perhaps not.

RP speakers are not easy to come by. Estimates suggest they amount to about 3-5% of the UK national population (summarised in Southern British English), and their social status means they are not all that accessible. Most of us would hardly ever hear RP spoken if it weren’t for the theatre, cinema, radio or television, which is where Deterding found his “similar to RP” informants. The five RP speakers presented here were chosen because public recordings are readily available. But it turns out that two of them have adopted RP, which need not matter provided acquisition was complete rather than partial. William Somerset Maugham’s first language was French, according to his Wikipedia biographer, although his parents were English and his father was legal adviser to the British Embassy in Paris; orphaned at 10 (1884), he stayed with his uncle in Whitstable (Kent), in Estuary English country; he boarded at King’s School in Canterbury, leaving around 1890 at 16 to continue his education at Heidelberg and then at St Thomas’s Hospital (London); getting teased for his English at school was an assumed cause for his stammer, and a measure of his progress towards RP. The other speaker with adopted RP is William Hague (politician), who grew up in Yorkshire; journalists have frequently teased him for his Yorkshire accent, but there’s scarcely any trace in the recordings I’ve heard; John Wells (1997, Our changing pronunciation, Transactions of the Yorkshire Dialect Society xix, 42-48) remarks on Hague’s STRUT vowel, so that will be checked; there was just one slip in the public address analysed here, when he missed the TRAP-BATH split, saying [vast] rather than [vɑːst] for vast.

Expressions like MOUTH, GOAT, LOT, THOUGHT, STRUT etc. are keywords used by John Wells (1982, Accents of English, C.U.P.) to denote lexical sets that participated in various dialect developments or sound changes (merging, splitting, shifting etc), and are useful for comparing accents.

Listen to these five RP speakers again, ordered by birth year:

William Somerset Maugham 1874, Harold MacMillan 1894, Ian Fleming 1908, Kingsley Amis 1922, William Hague 1961

Procedures

praatfftsliceFigure 1: Example of a narrow band spectral slice, showing the partial structure of a vowel to help determine which partials are included in a formant.
formantssberpwsmcFigure 2: F1/F2 vowel diagram for RP speaker William Somerset Maugham.

The recordings were all continuous speech, from interviews or public speeches, 3-5 minutes for each speaker. The online recordings had unfortunately been compressed by MP3 to low quality to facilitate handling on the web. Consequently broadband spectrograms and LPC formant tracking were not always distinct. Narrowband spectrograms turned out to be easier to work with, by identifying which harmonics were included in a formant, and which were not. Formants were sampled at moments where vowels were least affected by neighbouring consonants, as indicated by formant transition movements. The formants were then measured on the computer screen from spectral slices taken at the selected moments (Fig. 1). Stressed vowels in focally accented words were chosen, 3 to 5 instances of each phoneme depending on availability. There were few examples of the FOOT vowel, some speakers not providing any in their speech samples.

Based on narrowband spectrograms, measured from spectral slices.Figure 3: F1/F2 diagram for RP speaker Harold MacMillan
formantssberpifbFigure 4: F1/F2 diagram for RP speaker Ian Fleming
formantssberpkabFigure 5: F1/F2 diagram for RP speaker Kingsley Amis
formantssberpwhFigure 6: F1/F2 diagram for RP speaker William Hague

The F1/F2 diagrams for each RP speaker are shown at Figures 2-6. The frequency scales of the vowel diagrams are compressed to the Mel scale, so that equal linear intervals are perceptually equal in Mels (Stevens, Stanley, Volkman J, & Newman E. B., 1937, A scale for the measurement of the psychological magnitude pitch, Journal of the Acoustical Society of America 8, 185–190; Fant, Gunnar, 1968, Analysis and synthesis of speech processes, in B. Malmberg (Ed.), Manual of Phonetics, Amsterdam, 173-177). Other psychophysical scales are also found in the literature, a straight logarithmic transformation (yielding linear octaves and semitones), and Bark transformations (based on auditory critical bands).

harringtonqevowelsc
Figure 7. Results from Harrington et al. (2000) Journal of the International Phonetic Association 30: 63-78, the Queen’s Christmas broadcasts in the 1980s, lax vowels (left) and tense vowels (right).
formantssbedetmFigure 8: F1/F2 vowel diagram for the individual averages of Deterding’s five male speakers B C H J K. The LOT zone is split into blue and red areas (see text).
formantssbedetfFigure 9: F1/F2 vowel diagram for the individual averages of Deterding’s five female speakers A D E F G.

Examples from Harrington et al. (2000) are shown in Fig. 7, showing vowel diagrams from the Queen’s Christmas broadcasts in the 1980s (the authors had additional results from the 1950s and 1960s in their Fig. 1). Figures 8 and 9 show vowel diagrams for Deterding’s (1997) male and female informants respectively, based on data in his results tables.

20140206DGB01Figure 10: The effect of degree of constriction (Amin), and degree of mouth opening (A/l) at the four constriction locations for vowels. Superimposed for comparison is an F1/F2 diagram for a home counties SBE speaker (recorded from the radio in the 1970s). The ʌ zone refers to the STRUT vowel, ʌ being the traditional transcription (although this speaker had an [a]-like timbre).  From Wood, Journal of Phonetics 7, 25-43.

Since the 1950s, it’s been customary to interpret F1 frequency as degree of tongue lowering, and F2 as degree of tongue fronting, with reference to the Bell vowel model (The Bell vowel model, and The acoustic weaknesses of Bell’s vowel model). The true relationship of vowel formant frequencies to articulation is much more complex than Bell foresaw (see Interpreting vowel articulation from formant frequencies). For example, the FLEECE, KIT and DRESS zones seen in Figures 2-6 require the vocal tract to be narrowed along the hard palate; the GOOSE and FOOT zones require a narrowing along the soft palate; the THOUGHT zone requires a narrowing in the upper pharynx; the TRAP, STRUT, BATH and LOT zones require a narrowing in the lower pharynx (Sidney Wood, 1979, A radiographic analysis of constriction locations for vowels, Journal of Phonetics 7, 25-43). Figure 10 shows how varying the degree of constriction and mouth opening affects the frequencies of F1 and F2 at each of these four constriction locations. Figure 10 also includes a vowel formant diagram for a speaker of home counties SBE, for reference (the two cases don’t match perfectly because the speaker’s vocal tract was shorter than the model).

Comments

A quick comparison of the RP vowel diagrams in Figs. 2-9 and the the home counties SBE vowel diagram in Fig. 10 shows just how discrete any differences are between the two accents among the monophthong vowels (LOT and THOUGHT are [ɒ] and [ɔː] in RP, [ɔ] and [oː] in home counties SBE, respectively). Why should any differences matter? For the linguist, they provide criteria for discriminating between accents or dialects, and establishing directions of sound change in different speech communities. For the language teacher, SLP therapist, or dialect coach, they offer criteria for targets and for assessing progress. For a speech community, they signal group identity and membership.

The potential differences, as recorded by Daniel Jones (1918 and later, Outline of English Phonetics, Teubner, later Heffer) and A. C. Gimson (1962, An Introduction to the Pronunciation of English, Arnold), concern the THOUGHT (together with NORTH and FORCE), LOT, and STRUT vowels. More recent observations concern the GOOSE vowel.

  • THOUGHT.

Listen to examples of the [ɔː]-like THOUGHT vowel by the five RP speakers:

Three examples from each: often, forties, course; short, course, more; straightforward, war, taught; endorse, importance, more; lord, forebears, your

For comparison, listen to examples of the [oː]-like THOUGHT vowel by five home counties SBE speakers:

Three examples from each: performance, glory, warzone; absorbing, exhausted, dawn; thoughts, talk, endorsed; sports, important, forties; withdrawing, proportions, forces

    • Jones and Gimson describe the RP THOUGHT vowel as [ɔː], while the home counties SBE example (Fig. 10) shows [oː]. Wells (1982, vol. 2, p. 310) reports [oː] in popular London speech. On the formant diagrams, Figs. 2-10, [o] is closer to [u] (towards F1 around 400Hz), while [ɔ] is further away from [u] (towards F1 around 500Hz).
    • Gimson found [oː] increasingly common in Advanced RP (Wells’ U-RP), Wells (1982, vol 2, p. 293) dating this change to the past half-century without reference to any particular variety of RP. There were no examples of such a tendency from the five RP speakers.
    • The data of Harrington et al. (my Fig. 7 here) from the Queen’s Christmas broadcasts illustrates a shift from [ɔː] towards [oː]. The F1 range for her THOUGHT vowel shrank from 400~700HZ in the 1950s to 400~600Hz in the 1960s and 1980s, indicating an earlier wider [ɔː~oː] variation shifting towards mainly [oː] in the later years.
    • Deterding’s data for his male group is seen at Fig. 8. The THOUGHT zone had F1 around 400~500Hz, expected for the home counties [oː] timbre. However, speakers B and H were both at the higher F1 end of this zone, nearer the RP [ɔː] timbre. This might single out C J K as possible home counties candidates. His female group (Fig. 9) were clustered with F1 mostly below 400Hz, despite their shorter vocal tracts, which definitely points to a darker [oː] timbre, and leaves a large empty space between their LOT and THOUGHT zones. Perhaps they weren’t RP speakers after all. Or were they following Her Majesty’s lead in shifting their THOUGHT vowel from expected [ɔː] to [oː], closing yet another difference between RP and home counties SBE?
    • There’s been a consistent difference between [ɔː]-like RP THOUGHT and [oː]-like home counties SBE THOUGHT in the past, that has differentiated the two accents. However, that difference could disappear if coming generations of RP speakers consolidate the THOUGHT vowel shift from [ɔː] towards [oː].
  • LOT

Listen to examples of the [ɒ]-like LOT vowel by the five RP speakers:

Three examples from each, some repeated: honest, honest, competent; office, historic, economy; watch, foreign, job; neurotic, moral, stocks; honouring, colleage, policies

Listen to examples of the [ɔ]-like LOT vowel by five home counties SBE speakers:

Three examples from each: dots, wants, Hollywood; from, Johnson, comedy; not, gods, teleology; modern, lots, spin off; authority, strongest, crossing

    • Jones always transcribed the RP LOT vowel as /ɔ/, but it’s clear from his descriptions that the actual pronunciation was a rounded [ɑ], i.e. [ɒ]. Gimson adjusted the phonemic notation to /ɒ/. This timbre contains a hint of [ɑ] and Figs. 2-6 show the RP LOT zone very near the BATH zone, with F1 around or above 600Hz. Neither Jones nor Gimson mentioned any regional [ɔ]-like variants. In contrast, Fig. 10 illustrates the home counties SBE [ɔ]-like zone nearer the /o:/ zone, with F1 below 600Hz.
    • The Queen’s LOT vowel (Fig. 7) was always [ɒ]-like (F1 600~800Hz).
    • Taking Detterding’s male group first (Fig. 8), there were two clusters: B and H with F1 higher than 600Hz (blue zone), and C J K with lower F1 around 450~550Hz. This suggests B and H had the RP [ɒ] timbre, while C J K had the home counties [ɔ] timbre. The female group was closely clustered near the BATH zone with F1 600~700Hz, suggesting all had targeted the RP [ɒ] timbre. This singles out C J K as possible home counties SBE candidates.
    • There are no reports yet of RP shifting its [ɒ]-like LOT vowel towards the home counties SBE [ɔ]-like LOT.
  • STRUT.

Listen to examples of [a]-like STRUT by the five RP speakers:

One example from each, repeated three times: sometimes; country; publisher; eachother; hundred

Listen to examples of [a]-like STRUT by the five home counties SBE speakers:

One example from each, repeated three times: lucky; company; punishing; summer; structured

  • Daniel Jones (followed by Gimson) always described RP STRUT as an unrounded [ɔ], i.e. [ʌ]. Jones, and Gimson, recommended [a] as a best substitute if foreign learners fail to obtain the precise sound. Jones also observed that [a] “is actually used as a substitute for it in some English dialects (including London)”. Figure 10 confirms an [a]-like STRUT vowel for the home counties SBE speaker there. Strangely, Figures 2-6 show that the five RP speakers also pronounced STRUT as [a], Harold MacMillan and Ian Fleming slightly darker coinciding with the [ɑ] of BATH (differentiating STRUT from BATH by quantity). What appeared to be an expected difference ([ʌ]/[a]) between RP and home counties SBE just didn’t occur in any these samples. RP and home counties SBE speakers alike, the TRAP, STRUT and BATH vowels were all strung out along the bottom of the vowel diagrams, distinguished by their F2 frequencies: highest for TRAP, intermediate for STRUT and lowest for BATH.
  • The same distribution can be seen in the Queen’s data in Fig. 7. Her STRUT vowel was also [a]-like, between TRAP and BATH, with F1 as high as 600~1000Hz.
  • Deterding’s groups (Figs. 8 and 9) both exhibit the same result as all the previous data, an [a]-like spectrum for STRUT, between TRAP and BATH.
  • None of these RP speakers offered an [ʌ]-like example for STRUT, although it was given by Jones and Gimson for 60 years or more as the only RP pronunciation, with [a] occurring as regional or foreign learner substitutions. Nor does there seem to be any  consistent difference between the STRUT vowels of RP and home counties SBE, despite any expectations prompted by the standard RP literature. The apparent elusiveness of [ʌ] deserves thorough investigation.

 

  • GOOSE.
    • Jones always described the RP GOOSE vowel as having a more advanced tongue position than cardinal [u], and even more advanced after /j/. This means that this [u]-like vowel is not so dark as cardinal [u] (i.e. F2 would not be as low as 700 or 800Hz), and even brighter after /j/ (F2 perhaps around 1200Hz or more). Gimson followed Jones, adding a reminder that English does not have to respect an /u/-/y/ contrast, allowing more variability for RP /u/. Jones didn’t mention any regional variants, but Gimson reported that centralization is a characteristic of the regional dialect of London, adding that “considerable centralization amongst RP speakers is, therefore, inhibited to a certain extent for social reasons”. Jonathan Harrington et al. (2011, The contributions of the lips and the tongue to the diachronic fronting of high back vowels in Standard Southern British English, Journal of the IPA 41, 137-156) cite a number of reports from Englishes around the world, confirming a widespread brightening of the GOOSE vowel, including [y]-like allophones. Figures 2-6 show that those five RP speakers followed this same progression, darkest for those born around 1900 and steadily brighter through the 20th century.
    • The Queen’s Christmas broadcast data reported by Harrington et al. 2000 show her F2 in the GOOSE zone aproaching 2000Hz in the 1950s and exceeding 2000Hz in the 1960s and 1980s. That’s not quite so dramatic as it sounds if we allow a few hundred Herz for a shorter vocal tract, but the tendency is clear, there were some very bright instances of her GOOSE vowel.
    • Deterding’s data (Figs. 8 and 9) was averaged for each speaker, so we can’t see separate instances of high F2 for GOOSE. What can be seen is that F2 varied by as much as 500Hz across the individual averages of the male speakers, so that some had a much higher F2 than others.
  • TRAP.
    • Jones and Gimson agreed in their descriptions of RP TRAP, locating it just below cardinal [ɛ] in their diagrams. Wells reported a recent tendency to shift TRAP further away from DRESS and nearer [a], i.e. a change from an [ɛ]-coloured [æ] timbre to an [a]-coloured [æ] timbre. Evidence of this change can be seen in the diagrams for the five RP speakers (Figs. 2-6). William Somerset Maugham and Harald MacMillan, born in the 19th century, Figs. 2 and 3, have their TRAP zones very close to their DRESS zones. The other three, born in the 20th century, Figs. 4-6, have their TRAP zones further away from their DRESS zones.
    • Harrington et al. (2000) report the same tendency within the speech of the Queen’s Christmas broadcasts. They found her TRAP zone overlapping the DRESS zone in the 1950s, but TRAP was further away in the 1960s and 1980s (this cannot be seen in Fig. 7, which shows the 1980s data only).
    • Deterding’s male and female groups both exhibit two clusters, Figs. 8-9, one close to the DRESS zone (speakers D, F and C, H), and one close to the STRUT zone, reflecting the reported RP sound change.

Conclusions

RP and home counties SBE are certainly very close in an objective sense, for example it turns out that the only real differences today among the monophthong vowels are RP [ɒ] contra home counties [ɔ] for LOT, and RP [ɔ:] contra home counties [o:] for THOUGHT. But there was a time when the RP-speaking community were very much aware of that slight difference, and home counties SBE speakers were expected to modify it to RP for careers in most professions, including broadcasting. Shibboleths are typically based on subtle differences. Now, in contrast, RP is giving up that slight difference, step by step. Currently, RP [ɔ:] for THOUGHT is being shifted towards home counties [o:], leaving RP [ɒ] and home counties [ɔ] for LOT as the only potential difference in the future.

All the examples reviewed here exhibited an [a]-like timbre for the RP STRUT vowel, rather than the [ʌ]-like timbre expected from standard descriptions.

All the examples reviewed here confirm that the former dark rendering of RP GOOSE, with F2 rarely going higher than 1000~1200HZ, has given way to much brighter [ʉ]-like timbres, and at times even [y], with F2 as high as 1500~2000Hz.

All the examples reviewed here confirm a former rendering of RP TRAP that was nearer DRESS, while later generations have shifted it away from DRESS towards STRUT.

The vowels of the five home counties SBE speakers, heard in the recordings above, will be reported separately. The elusive [ʌ]-like timbre for STRUT, and GOOSE fronting, will also be looked at more closely.

Top
©Sidney Wood and SWPhonetics, 1994-2014

______________________________________________________________________

Advertisements