PhonaestheticallyspeakingDAVID CRYSTAL

An examination of why people regard

some words as inherently more beautifulthan others

EVERYnow and then, people ponder about themost beautiful words in the English language ­beautiful, that is, in terms of sound rather thanmeaning. The study of the expressive proper­ties of sound is called phonaesthetics. Here is aselection of phonaesthetic opinions.

• From Willard Funk, the lexicographer: tran­quil, murmuring, mist, chimes, dawn, hush,luminous, lullaby, golden, melody. Pressed toextend this list, he added 21 other items beforecalling a halt: chalice, jonquil, thrush, marigold,myrrh, damask, asphodel, oleander, halcyon,oriole, fawn, anemone, tendril, amaryllis,camelia, ceulean, rosemary, alyssum, gossamer,mignonette, bobolink.

• From poet John Kitching (in no specificorder): velvet, melody, young, gossamer, crystal,autumn, peace, mellifluous, whisper, tranquil,lace, caress, silken, willow, mellow, lullaby,dawn, shimmer, yellow, silver, marigold, golden,dream, harmony, olden, blossom, champagne,sleep, dusk, magic, hummock, love, mist, dar­ling, laughter, butterfly, charity, eiderdown, sky,parakeet, rosemary, froth, gazebo, ivory, syl­labub, vacillate, mesmerism, echo, fate,jacaranda, harlequin, chrysalis, violin, enigma,tart, sycamore, pomp, chinchilla, truffle, myrrh,bewildered, claret, akimbo, fur, flamingo,celandine, ominous, tantalize, wine, antimacas­sar, jewel, skill, russet, buckram, delight, thrill,clavichord, didgeridoo, doppelganger, fractious,zoo.

• In a 1980s' newspaper report, an unnamed

novelist is said to have chosen: peril, moon,shadow, azure, carnation, heart, silence, forlorn,April, apricot.

• In a 1980 Sunday Times reader's poll, melodyand velvet tied for first place; third was a tiebetween gossamer and crystal, followed byautumn, peace, tranquil, twilight, and murmur,with caress, mellifluous, and whisper tying fortenth place.

It is impossible, of course, to separate soundand meaning totally: the writer who foundperil a beautiful word is actually doing some­thing rather unusual, in being able to disassoci­ate sound from meaning so radically. On thewhole, pleasant-sounding words have positiveand desirable meanings, or represent favouredsemantic domains, such as birds and flowers.

DAVIDCRYSTALread English at University CollegeLondon, and has held posts in linguistics at theUniversity Collegeof North Wales, Bangor, and atthe University of Reading, where he taught fortwenty years. He works currently as a writer,lecturer, and broadcaster on language andlinguistics. He is the editor of 'LinguisticsAbstracts'and 'ChildLanguage Teachingand Therapy'. Amonghis more recentpublications are 'Listen to YourChild~ 'Who CaresAbout English Usage?','LinguisticEncounters with Language Handicap', 'TheCambridgeEncyclopedia of Language', 'TheCambridge Encyclopedia',and, most recently,'The CambridgeEncyclopedia of the EnglishLanguage'.

8 English Today 42, Vo!. 11, No. 2 (April 1995). Copyright © 1995 Cambridge University Press

But, we may still ask, out of the thousands ofwords in English with positive and desirablemeanings, why have these particular ones beenchosen? Several turn up on more than one list(autumn, caress, crystal, dawn, golden, gos­samer, lullaby, marigold, mellifluous, melody,mist, myrrh, peace, rosemary, tranquil (3times), velvet, whisper). Is there somethingabout the phonetics of these words whichmakes them so attractive? Which phonologicalpatterns are most involved?

It is a useful exercise to stop at this point, andjot down the sound effects which strike you asbeing particularly important, in the above lists,before comparing your list with the results ofthe analysis below. The task is to notice notonly which sounds and sound patterns are fre­quently used, but also - rather more difficult ­which are not used at all. 'Ugly' words, pre­sumably, would use these 'missing' soundsrather more often.

For this exercise, I made a phonological pro­file of the distribution of the sounds in thesewords, marking where in a syllable a soundoccurred, the number of syllables in the word,and where the stress fell. Such a profile enablesthe analyst to see at a glance which sounds areused, and how often, and which sounds are notused at all. I used the analysis of the Englishsound system as it is presented in A C Gimson'sIntroduction to the Pronunciation of English(Edward Arnold, 5th edition, 1994), andassumed Received Pronunciation. Here are themain findings. (Phonetic symbols are glossedin Panel 1.)

Consonants

• The consonants clearly divide into twotypes: high frequency and low frequency. Justeight items account for 73% (274) of all conso­nants: /V is top, with 59 instances (16%), fol-

Phonetic symbols ~

Conversation

% of all PhonaestheticN/% of all% of allsegments

wordssegments (625)consonants (377)

Consonants n

7.58 I59/9.4415.65t

6.42 ill40/6.4010.61d

5.14 s35/5.609.28s

4.81 n33/5.288.75I

3.66 r29/4.647.69Cl

3.56 k28/4.487.43r

3.51 t26/4.166.90ill

3.22 d24/3.84 6.37-------------------k 3.09 f12/1.92 3.18w

2.81 b11/1. 762.92z

2.46 p10/1.602.65v

2.00 v811.282.12b

1.97 IJ8/1.282.12

f1.79 w8/1.282.12

P

1.78 g7/1.121.86h

1.46 z7/1.121.86IJ

1.15 J7/1.121.86a

1.05 h6/0.961.59bJ

0.96 g5/0.801.33J

0.88cB5/0.801.33

cB

0.60 J5/0.801.33g

0.41 e3/0.480.80e

0.37 31/0.160.263

0.10 Cl00

PHONAESTHETICALLY SPEAKING 9

Vowels

g 10.748.33e

2.97aJ

1.83A

1.75eI

1.71I:

1.65

gU

1.51

<e

1.45D

1.37

::J:

1.24u:

1.13

U0.86

a:0.79

au0.61

3:

0.52

£g0.34

Ig0.21

::JI0.14

Ug

0.06

N/%ofall

% of all vowels~

segments (625)

(248)

g

61/9.76 24.604917.84

19.76<e

24/3.84 9.68e

1412.24 5.65I:

14/2.24 5.65al

1312.08 5.24gU

13/2.08 5.24A

12/1.92 4.84D

10/1.60 4.03eI

8/1.28 3.23u:

7/1.12 2.82::J:

7/1.12 2.82a:

7/1.12 2.823:

4/0.64 1.61Ig

2/0.32 0.81aJg

1/0.16 0.40au

1/0.16 0.40Ug

1/0.16 0.40

lowed by Iml (40), Isl (35), 1nl (33), Irl (29),Ikl (28), It! (26), and Idl (24). If this rankingis compared with that found in conversation(see Panel 1), the use of III and Iml is note­worthy.

• There is then a big jump before reaching thelow-frequency consonants: If! (12), Ibl (11),Ipl (10), Ivl (8),/1]1 (8), Iwl (8), Igl (7), Izl(7), J (7), Ihl (6),1\1'1 (5), letl (5), Ijl (5),/81(3),131 (1). 101 is the only consonant whichdoes not occur at all.

• If we group these consonants into typesaccording to their manner of articulation, fric­tionless continuants are commonest (182: 101oral, 81 nasal), followed by plosives (106),fricatives (79), and affricates (10). As there areonly four oral continuants ClI, r, w, j/) andthree nasals Clrn, ll, 1]/), but six plosives Clp, b,t, d, k, g/) and nine fricatives CIf, v, 8,0, s, z, f,3, hi), this distribution is noteworthy. Continu­ants are definitely popular.

• If we group consonants in terms of theirplace of articulation, there is very little differ­ence from everyday conversation. Front(notably labial) consonants are a little morefrequent in phonaesthetically pleasing words,

and back ones (velar and glottal) a little less so.

• Only 61 (16%) of the consonants appear asadjacent consonants (e.g. Idr-, -Id-I. There areonly three cases where three consonants cometogether (as in tendril). Many people have aninitial impression that consonant clusters arean important feature of phonaesthetic words:in fact, they are not.

• In any word, is there a likelihood that conso­nants influence each other in sequence? If aword contains an Iml, will there be anotherIml following in the word? This does not hap­pen much. In only 6 cases are there instances ofa consonant being immediately followed byexactly the same consonant (whether sepa­rated by a vowel or not) (e.g. bobolink); five ofthese, interestingly, involve plosives.

• What about grouping sequences of conso­nants into place of articulation categories (e.g.labial, alveolar, velar, glottal)? If a word con­tains an Iml, will there be another bilabial con­sonant following in the word? There is little ofinterest here. There is a slight preference(60:40%) in favour of changing the place ofarticulation from one consonant to the next,but nothing very dramatic.

10 ENGLISH TODAY 42 April 1995

• What about grouping sequences of conso­nants into manner of articulation categories(e.g. plosive, fricative, nasal, continuant)? If aword contains an Iml, will there be anothernasal consonant following in the word? Herethere is a major preference (84:16%) in favourof changing the type of consonant. A wordapparently sounds prettier if the manner ofconsonantal articulation changes as the sylla­bles pass by.

• The voicing pattern within a word seems tobe of little interest. Only 132 (35%) of the con­sonants are voiceless, but this is almost exactlythe proportion we would expect in everydayconversation. This carries through into theword patterns: if a consonant is voiced, there isa 60% chance that the next consonant is also

going to be voiced; if a consonant is voiceless,there is a 30% chance that the next one is alsogoing to be voiceless.

Vowels

• Of the 172 vowels, the unstressed vowel Iglis commonest (61), showing that words ofmore than one syllable are preferred. Of the 114words, only 25 are monosyllabic, in fact; thelargest category (45) is words of three or moresyllables. Of these, 29 (65%) have the chiefstress on the first syllable .

• The other common vowel is /II (49), whichoccurs over twice as often as the next vowel

1a;1 (24), followed by lel, li:/, lail, etc. (seePanel 1b). This is close to the vowel rankings ofeveryday conversation. The only pure vowelwhich is not used at all is lul (as in put). Peo­ple have the impression that long pure vowelsare important in phonaesthetically pleasingwords: in fact, four of the five long vowels arenot even in the top ten of the vowel list. Diph­thongs do better.

• If we look at vowels in terms of where theyare articulated in the mouth - whether at the

front (or, in the case of diphthongs, starting atthe front) (e.g. la;, ar/), central (e.g. /g, gu/),or back (e.g. lu:, JI/), we find an interestingpattern. There are more front vowels here thanin everyday conversation (20% vs. 15.5%).There is also a strong tendency for a polysyl­labic phonaesthetic word to start with a vowelin the front position (in 52 cases; 58%) andthen either to continue using front vowels laterin the word (e.g. camelia) or to have these

PHONAESTHETICALLY SPEAKING

Caprice in consonantsSome words have such a lovely soundIt's pleasant to roll them round and roundAnd savor their syllables on the tongue, ­Words like oriole, melody, young.

Other words, though, of ungracefulletter,Harsh, abrasive, ... sound even better!These are words of intrinsic beaury, ­Service, conscience, kindliness, dury.

- Alma Denny, New York

other vowels move in a backwards direction

(e.g. apricot; 40 cases; 45%). There are nocases of words in which all the vowels are

found in the back position; and there are only10 instances (12%) where the word beginswith a back vowel and has later vowels further

forward (e.g. autumn). Similarly, there areonly 13 cases (15%) where the word beginswith a central vowel, with later vowels eitherstaying central or moving further forward (e.g.murmuring). There is no real trend for a wordof three syllables or more to have its vowelsspread between front, central, and back (e.g.sycamore), a pattern found in only 13 out of 45cases (29%).

• If we look at vowel articulation in terms of

whether the vowels are high, mid, or low in themouth, there is less to say. There are no realdifferences here compared with the propor­tions found in everyday conversation. The onlystrong tendency is for a polysyllabic word tobegin with a vowel in mid or low position (66cases; 76%) and for later vowels in the word tomove in an upwards direction e.g. ivory; 26cases, 58%). Only 12 cases (27%) move in theopposite direction, starting high, and endingup low (e.g. syllabub).

If these trends are respectable, given the smallsample, it is possible to see how we can createphonaesthetically pleasing new words. Itwould seem advisable to give them three sylla­bles, to stress the first syllable, to use at leastone Iml or III (preferably both), to introducehigh-frequency consonants and avoid low-fre­quency ones, to have at least three differentmanners of consonantal articulation, to keepthe vowels short, and to have the vowels movefrom mid towards high, and from front towardsback.

11

A matrix of criteria 2

3+StressUsesUsesUseofNo use3+ diffOnlyFront>Low>sylls

on 1stImlIVotherof lowmannershortCentrelMidisyll

highfreq Csof articVsBackHighfreq Cstremulous

jjjjjjjjjjalyssum

jjjjjjjjjXalumnus

jxjjjjjjjjramelon

jjjjjjjjjjdrematol

jjjjjjjjjjPimlico

jjjjjxjxjxWapping

xjxxxxjjXjphlegmatic

jXjjjXjjjXflatulent

jjXjjXjjXXgripe

Xn.a.XXjXXXn.a.n.a.

jazz

Xn.a.XXXXXjn.a.n.a.

tartXn.a.XXjjXXn.a.n.a.

zoo

Xn.a.XXXXXXn.a.n.a.

n.a. = not applicable

A matrix of criteria

We can see this more clearly if we construct amatrix with these ten criteria across the topand candidate words down the side (see Panel2). Tremulous, from the above listings, scores10 out of 10, alyssum 9. On the other hand, sev­eral of the words in these listings have very lit­tle going for them: tart scores 2; zoo scoreszero! Words which have scores of less than

about 6 are probably in a list chiefly because oftheir meaning rather than their sound.

And ugly words? We can of course use thematrix to test the opposite effect. A group ofelocutionists once produced a listing of whatthey considered to be the ten ugliest-soundingwords in the language: flatulent, treachery,crunch, phlegmatic, sap, jazz, plutocrat, cacoph­ony, gripe, plump. Are these genuinely unpleas­ant in sound, or is it the meaning which is theroot of the objection? The phonaestheticmatrix shows that some of them are actuallyquite pleasant-sounding, really. Phlegmaticscores 7 out of 10, and flatulent 8. On the otherhand, there seems to be very little going forgripe and jazz (both lout of 10).

Using this approach, we could probably

extend our list of pretty words quite substan­tially, to include such items as emulate andalumnus. We could also start thinking uppretty words which do not yet exist, such asramelon and drematol (though of course suchword-forms may already be used to make nice­sounding names for pharmaceuticals andother products). The matrix also suggests why,if we wanted to write a romantic poem about,say, London Underground stations, our listwould very likely include Pimlico and Colin­dale, but exclude Goodge Street and Wapping.And why friendly space aliens receive suchnames as Alaree and Osnomian, why enemynames include Vatch and Triops, and why Klin­gons are likely to be just a tad less aggressivethan Kryptons. lE[j

Acknowledgement

The phonological profiling procedure isadapted from D. Crystal, Profiling LinguisticDisability (London: Whurr, 2nd edn, 1992). Iam also grateful to V. J. Cook of the Universityof Essex for a sight of his collection of aliens'names.

12 ENGLISH TODAY 42 April 1995