PHONETICNOTATIONSYSTEMS

JoVerhoeven&A.HirsonCity,UniversityofLondon,DivisionofLanguageandCommunicationScience,

PhoneticsLaboratoryThisisaslightlyupdatedversionofanarticlewhichwaspublishedas:Verhoeven,J.,Hirson,A.(2009).Phoneticnotationsystems.INJ.Verschuren(Ed.)HandbookofPragmatics,vol.13,pp.1-14.1.INTRODUCTIONPhonetic notation systems have long been used as an essential tool in thephoneticdescriptionoflanguagesoftheworldandalsomoregenerallyinotherbranchesof the linguistic sciences. In this article, oneof themostwidelyusedphonetictranscriptionsystemswillbediscussedindetail,i.e.thealphabetoftheInternationalPhoneticAssociation.Thisalphabeticsystemis firmlyrooted inastronginternationalphonetictraditionwhichemergedinthelastdecadesofthe1800s.Beforegoingintothistranscriptionsystem,however,itmaybeusefultodefine the subject area of phonetics, which is done elegantly in Abercrombie(1967)bymakingarigorousdistinctionbetweentheconceptsof languageandmedium.Distinguishing language frommedium isdistinguishingpatterns fromtheirmaterialembodiment.“Language,wecouldsay,isform,whilethemediumis substance” (Abercrombie 1967: 1). In other words, language refers to theunderlying abstract patterns and rules of the language,whilewords, languagerefers to the underlying abstract patterns and rules of the language,while themediumactsas itsvehicle, i.e. itexternalizes language insomeformmediatingbetween ‘speaker’ and ‘listener’. Linguistics is the discipline that studieslanguage, while phonetics can be appropriately regarded as the study of themediumofspokenlanguage:“Thestudyofthemediumofspokenlanguage,inallitsaspectsandallitsvarieties,constitutesthesubjectofphonetics.Phoneticsisconcernedwiththemediumasusedinspeakingallhumanlanguages(…),andasused in all styles of speech (whether supposed to be good or bad, normal orabnormal)”(Abercrombie1967:2).Abercrombieclassifiesthedifferenttypesofmedia on the basis of the reception activity involved. In spoken language themedium is aural i.e. perceived by the listener’s ear. The production activityinvolvedistheuseofthearticulatorstoproducespeechsoundstobeperceivedbyalistener.Inwrittenlanguagethemediumisvisualorperceivedbytheeyeofareader.Inthiscase,themediumistheproductoffineco-ordinatedmovementsof hands and fingers to produce shapes on paper. Finally, there is the tactilemediumperceivedviathesenseoftouchofthereceiver.Agoodexampleofthisis theBraille system inwhich specificpatternsofdotsareembossedonpaperand‘read’byfeelingthesepatterns.Atitsmostgenerallevel,thetermnotationreferstotheconversionprocessofonemediumintoanother.Phoneticnotationspecifically transforms the aural medium into a visual or tactile medium toprovideanaccurateandpermanentrecordofthewaysinwhichspeechsoundsareproducedinlanguages(articulation).Furthermore,suchrepresentationalsoindicateswhat these speech sounds sound like (perception).Phoneticnotationthus informs both about the speech production and the speech perception

processesinvolvedintheauralmedium.2.HISTORICALTAXONOMICPERSPECTIVEMany attempts have beenmade in the past to develop phonetic transcriptionsystems.Abasicdistinctionthatistraditionallymadeisbetweenalphabeticandanalphabetic notation systems. In analphabetic systems, individual speechsounds are represented as strings of symbols in which each symbol explicitydescribesaspecificarticulatory featureof thesound involved.Thebest-knownexamples of such phonetic transcription systems are those developed by theDanish linguist and co-founder of the International Phonetic Association OttoJespersen (1889) and the American phonetician Kenneth Lee Pike (1943).Jespersen’stranscriptionsystemsusescombinationsofGreekandRomanletters,numerals, italics, heavy type and, subscript and superscript letters. The Greekletters represent the active articulators involved. Active articulators are thosethatmoveduringspeechproduction.InJespersen’ssystemαreferstothelips,βtothetipofthetongue,γtotheuppersurfaceofthetongue,δtothevelumorsoftpalateincludingtheuvula,εtothelarynxcontainingthevocalfoldsandζtothe organs of respiration. The Roman letters refer to the passive articulators:these articulators are relatively passive and primarily capture place ofarticulation:labial(a–c),interdental(d),dental(e),alveolar(t),postalveolar(g),palatal(h, i),velar(j),uvular(k),pharyngeal(l).Doublingaletterorboldprintrefers to a broader place of articulation rather than a single point, while twosimultaneous places of articulation (double articulations) are represented bytwo letters separated by a comma. The numerals, finally, give a conventionalindicationof thedegreeof constrictionof the vocal tract createdby the activeand passive articulator: 0 = complete closure, 1 = narrow approximation etc.Thesesymbolscanbeusedtocreaterelativelycompactbutintractableformulassuchasα7ßgγ7δ0ε1todenotethevowelinEnglish‘all’.Pike’s analphabetic transcription system developed in Phonetics (Pike 1943)providesevenmoredetailaboutspeecharticulationthanJespersen’sandlookseven more Byzantian by today’s standards. His ‘functional analphabeticsymbolism’ consists of Roman and italic letters in upper and lower case. Theitalic symbols specify the broad categories of the articulatory mechanismsinvolved,whiletheRomanheadingsexpresssubclassifications.Inhisnotationalsystemthe transcriptionof thevoicelessalveolarstop [t] requires34symbols:MaIlDeCVveIcAPpaɑtdtltnransfsSiFSs. In this ‘formula’, the energy-producingmechanismofthesoundis for instancespecifiedbythesymbols following left-mostM:itindicatesanairstreammechanism(a)inwhichtheinitiator(I)isthelungs(l).Inaddition,thedirectionoftheairstream(D)isegressive(e).Both systemsdescribedabovehave in common that the symbols are arrangedhorizontallywithafront-to-backarticulationorganisationrepresentedfromleft-to-right.Theygraphicallyillustratethatprovidingatrulyexhaustivearticulatorydescription of speech sounds is impossible: Pike has argued for instance thateven his own degree of articulatory detail is “suggestive but by no meansexhaustive”(Pike1943:153).Whethercompleteorincomplete,theyarehighlyimpractical in everyday use. In Pike’s system for example, the transcription ofsingle speech segments such as the alveolar nasal [n] and the half-close back

vowel[o]require79and88symbolsrespectively. It isnothardto imaginetheexplosionofinformationinvolvedinthetranscriptionofsimplewords,letalonelonger stretches of speech. This is probably the reason why the practicalimplementationofthesetranscriptionsystemsnevertookoff.Whereasanalphabeticnotationsystemsrepresentspeechsoundsbyasequenceofsymbols,alphabetictranscriptionsystemsaimtorepresentindividualspeechsoundsbysinglesymbolsasmuchaspossible.Withinsuchnotationsystems,adistinction is generallymade between Roman-based systems and non-Roman-based (or iconic) systems. In iconic systems the symbols are not chosenarbitrarily, but they are intended to somehow resemble what they represent.ThiscanbeillustratedwithreferencetoAlexanderMelvilleBell’sVisibleSpeechwhich dates back to 1867. In this system the symbol O denotes open glottiswhich“ispictorialoftheexpandedbreath-channelinthethroat”(Bell1867:46).This symbol is contrastedwith Iwhich stands for “glottis closed for vocal foldvibration“.Furthermore,alltheconsonantsymbolsaresimilaringeneralshape(avariantofC),buttheorientationoftheopeningdiffersdependingontheplaceofarticulationinvolved.Oneoftheadvantagesoficonicsystemsisthattheyarequite explicit about the articulatory dimensions involved, but their maindisadvantageisthattheycanbedifficulttoreadsincemanyofthesymbolsarevery similar in appearance. Furthermore, they can be difficult to rememberunless frequently used. Abercrombie (1967) considered their biggestdisadvantagetobethatthesymbolsareoftenmotivatedbyaparticulartheoryofphonetic description, although arguably this objectionmay be directed at anysystemforphonetictranscription.Roman-based notation systems differ from iconic systems in that the symbolschosen to represent speech sounds are chosen arbitrarily. Probably the mostwidely-used and most generally-accepted system of Roman-based phoneticnotation was developed by The International Phonetic Association or IPA, anorganisation which was inaugurated in 1886 as The Phonetic Teachers’AssociationbyagroupofFrenchlanguageteachers.3.THEIPAALPHABETThephonetictranscriptionsystemoftheInternationalPhoneticAssociationisanalphabeticnotationsystem.Thismeans that itusesasetof symbolswhicharebasedonRomanorthography,supplementedinvariouswaysbyelementsfromother symbol sets which are often modified slightly to harmonise with theRoman-based phonetic symbols. In such alphabetic systems, each phoneticsymbolrepresentsacompositesetofarticulatoryfeatures.E.g.[p]representsa‘voiceless bilabial plosive’ powered by a pulmonic egressive airstream, i.e.informationaboutanyparticularsoundrelatestotheairflow,whetherthevocalfoldsarevibrating(voiced/voiceless),theplaceofarticulation,andthetype(ormanner)ofarticulation.Theoriginaldesignprinciplesunderlying thisnotationsystemwerestatedinPrinciples(1948):1.Thereshouldbeaseparateletterforeachdistinctivesound;thatis,foreachsoundwhich,beingusedinsteadofanother,inthesamelanguage,canchangethemeaningofaword.

2.When any sound is found in several languages, the same sign should beusedinall.Thisappliesalsotoverysimilarshadesofsound.3.Thealphabetshouldconsistasmuchaspossibleoftheordinary lettersoftheromanalphabet,asfewnewlettersaspossiblebeingused.4.Inassigningvaluestotheromanletters,internationalusageshoulddecide.5.Thenewlettersshouldbesuggestiveofthesoundstheyrepresent,bytheirresemblancetotheoldones.6. Diacritic marks should be avoided, being trying for the eyes andtroublesometowrite.(Principles1948:3ofcover).

AlthoughtheIPAsystemisasystemforphoneticnotation,itisclearlyrootedinaphonological perspective: its most basic principle is the relevance ofphonological distinctiveness in assigning separate symbols to sounds whichfunctioncontrastivelyinlanguagesoftheworld.3.1CONSONANTSYMBOLSOFTHEIPAThe most up-to-date version of the IPA consists of 79 discrete ‘alphabetic’phoneticsymbols forconsonants.Thebiggestsetconsistsof69symbolswhichrefertopulmonicconsonants,i.e.consonantswhicharepoweredbyapulmonicegressiveairstreammechanism.Inthesesounds,theairstreamisinitiatedbytherespiratorysystem(thelungs).Therelativelylargenumberofsymbolsinthissetas compared to other sets is a natural reflection of the fact that 94% of thesounds in the languages of the world are made on a pulmonic egressiveairstreammechanism. These symbols are presented on various charts. In themain chart in Figure 1, the consonants are ordered in cells resulting from theintersectionofrowsandcolumns:therowsrepresentthedifferent‘manners’ofarticulation, while the columns represent the different ‘places’ of articulation.Themanners areordered frommore constrictedarticulations (upper rows) tomoreopenarticulations(lowerrows).

Figure1:IPAchartwiththesymbolsforpulmonicconsonantsThe sounds involving the strongest constrictionof the vocal tract areplosives,nasals,trillsandtaps/flaps.Fromanarticulatorypointofview,allthesesoundsinvolveat least inpartastrictureofcompleteclosure inthevocal tractsothattheairstreamispreventedfromescapingviathemouth.Inaplosivesuchas[d],

thetipofthetonguetouchestheroofofthemouthjustbehindtheupperteeth.Thistrapstheairstreaminsidethemouthsothatpressurebuildsupbehindtheocclusion.Uponreleaseof thecontactbetweenthearticulators, theairescapeswithabriefexplosiveburstwhichischaracteristicfortheperceptualimpressionof thesesounds.All the languages in theworlddescribed todate (100%)haveplosives.Thearticulationofnasalssuchas[n] issimilartothatoftheplosive[d] inthatthere is a complete obstruction of the airstream in the mouth created by thetongue tip touching the roof of the mouth just behind the upper teeth. Thisprevents the airstream from escaping via the mouth. In nasals, however, thevelum is loweredso that theairstreamisdirected into thenosealongwhich itescapes.96%ofthelanguagesoftheworldhavenasalspeechsegments.Intrillssuchas[r]thetonguetiptrillsagainstthepassivearticulatorandassuchthey are characterized by a sequence of intermittent complete closures of thevocal tract which result from aerodynamic conditions similar to vocal foldvibration.Trillsoccurin36%ofthelanguagesoftheworld.Tapsorflapsarecharacterizedbyaballisticactionofthetongue.Intaps,thetipof the tongue is raised from its neutral position and is thrown in a ballisticmovement against the roof of the mouth to make a very short closure whichinterruptstheairstream.Inflaps,thetipofthetongueiscurledbackinsidethemouth and then thrown forward back into its neutral position. During thismovement, the tongue flaps against the roof of the mouth to momentarilyobstructtheairstream.Bothtapsandflapsaredifferent fromstops intermsofthe ballistics involved and the inherently short duration of the closure. Thesesoundsoccurin35%ofthelanguagesoftheworld.Fricativessuchas[f],[s]and[ʃ]aremadeonastrictureofclose(i.e.incomplete)approximation: this means that the vocal tract is narrowed to a considerabledegreebythearticulatorsinvolved.Duringthearticulationofthesesounds,theairstreamisforcedthroughtheremainingnarrowgap,whichcausesanaudiblehiss.93%ofthelanguagesintheworldhavefricatives.Inlateralfricatives,thegapislocatedsidewayssothattheairescapeslaterally.The approximants are produced on a relatively opription of the IPA places ofarticen stricture: the gap between the articulators is wider than in fricatives,whichresultsinaminimalobstructionoftheairstreamsothatnoturbulenceorhissisgenerated.96%ofthelanguagesintheworldhaveapproximants.ThisIPAcategoryisfurthersubdividedintocentralandlateralapproximants.Thelateralapproximantssuchas[l]arecharacterizedbyplacingthetongueasanobstacleagainstthemedialpartoftheroofofthemouthsothattheairisforcedtoflowaroundit.In passing, it should be noted that some of the IPA terminology referring tomanner of articulation clearly reflects a perceptual approach to phoneticclassification, rather than a purely articulatory one. The term ‘plosive’ forexample elegantly captures the perceptual impression of explosion of thesesounds upon release of the articulators and in this sense it is different from a

term like ‘stop’ which refers more to the effect on the airstream of completevocal tractconstriction.Theeffectof thisperceptualorientation is that the IPAclassification sometimes highlights the perceptual differences between soundclasses rather than articulatory similarities. This is not always useful andmayobscureinterestingsimilaritiesbetweensoundcategories.Forexample,theIPAsystemtreatsplosivesandnasalsastwoquitedistinctsoundcategories.Thereisno doubt that the perceptual difference between these categories is quitesubstantial. Nevertheless, such classification obscures the fact that from anarticulatory point of view plosives and nasals are actually very similar in thatbotharemadeonastrictureofcompleteclosureintheoralcavity(Laver1994).Apart from the different constriction types, the IPA distinguishes between 11placesofarticulationwhichareorderedinFigure1fromfrontarticulationssuchasbilabial(totheleft)tobackarticulationssuchasglottal(totheright).Theseplacesofarticulationaredefinedbyspecifying thepassivearticulator involvedandareillustratedinFigure2:

Figure2:illustrationoftheIPAplaceofarticulationlabels.Theindividuallabelsare explained in Table 1, in which the percentages refer to the frequency ofoccurrenceofthecorrespondingsoundsinlanguagesoftheworld.Table 1:descriptionof the IPAplacesofarticulation.Occurrence indicates thepercentageof451UPSIDlanguageswiththisplaceofarticulation.

Place Description Occurrence

Bilabial orlabial

The articulators are the lower and upper lip forsoundslike[b]and[m]in‘bat’and‘mat’.

99.78%

Labiodental The lower lip approaches the upper teeth, forsoundslike[v]in‘veal’.

45%

Dental The tungue tipapproaches theupper teethas for[ð]in‘this’.

35.03%

Alveolar The tongueblade approaches thebony ridge justbehind the upper teeth. This creates sounds like[t]asin‘tell’

63.6%

Postalveolar The tongue blade moves towards the areabetween the alveolar ridge and the hard palate.Thisgivesrisetosoundslike[ʃ]in‘shield’.

64.08%

Retroflex The tipof the tongue iscurledbackand is raisedtowards the hard palate. The tongue typicallyadoptsaconcaveshape.

20.8%

Palatal Themidsectionofthetongueisliftedtowardsthehardpalate.Thetonguetypicallyadoptsaconvexshape.

89.0%

Velar The back of the tongue approaches the velum asfor[g]in‘go’.

99.6%

Uvular Thebackofthetongueisraisedtowardstheuvula,suchasinauvulartrill[R]

18.0%

Pharyngeal The tongue root moves towards the pharyngealwall

4.21%

Glottal Involves a constriction of the glottis (e.g. glottalstop[ʔ])

77.50%

In terms of the use of IPA symbols, it can be noted that for all manners ofarticulation except fricatives, there are no separate symbols for dental andpostalveolar places of articulation and as such these sounds are leftunderspecified:However, the full specifications can easily bederived from thealveolarsymbolsbymeansofdiacritics(seefurther).AthirdorganisationalprincipleunderlyingthismainconsonantchartinFigure1is thatwithin each cell of the table, voiceless segmentswith the samemannerand place of articulation are conventionally positioned left, while voicedsegmentsappearon the right. Inbroadoutline, voiceless soundsareproducedwithout vibration of the vocal folds, while voiced sounds do have vocal foldvibration. It is known that all languages of the world use voicing distinctionslinguistically,buttheproportionofvoicelessandvoicedspeechsoundsinsoundinventoriesmayvarysubstantially.Catford(1977)hascalculatedthevoiced-to-voiceless ratio for various languages. In European languages the number ofvoiced speech sounds tends to be substantially bigger than the number ofvoicelesssounds,givingrisetorelativelyhighvoiced-to-voicelessratios(French:3.55; English: 2.57), while the reverse is true for an Asian language such as

CantonChinese(0.70).Finally, it shouldbementioned that thechart inFigure1containsasignificantnumber of shaded and empty cells. The shaded cells on the chart representsoundcategorieswhicharedeemedphysiologicallyimpossible,whiletheemptyslots refer to sound classeswhich do not have separate symbols. However, inmany cases the symbols for these sounds can be easily derived by addingdiacriticstoexistingsymbols.AgoodexampleofthisisthebilabialapproximantwhichoccursinintervocalicpositionsinSpanish:thesymbolforthissoundcanbederivedbyusingthealphabeticsymbolforthevoicedbilabialfricative[β]andadding a diacritic to indicate that themanner of articulation ismore open, i.e.approximant.Hence,thesymbol[β̞]isobtained.Besides the symbols for pulmonic consonants, the IPA has 11 symbols torepresentconsonantsproducedonairstreammechanismsotherthanpulmonic:thesearelistedinaseparatetablewhichisillustratedinFigure3:

Figure3:IPAchartwiththesymbolsfornon-pulmonicconsonantsEjectivesarepoweredbyalaryngeal(glottalic)egressiveairstreammechanism.The initiator of this airstream mechanism is the larynx: the vocal folds areadducted to close the glottis and the larynx as awhole is pulledupwards in apiston-like action.This createspositivepressure above thevocal folds and theairflowsoutward.Ejectivesoundsareindicatedbyaddinganapostrophetothecorrespondingpulmonicsymbol.Itshouldbenotedthatejectivesneedarathertightconstrictionofthevocaltractsothatonlyplosives,fricativesandaffricatescan be changed into ejectives by means of a laryngeal airstream mechanism.Ejectives are by definition voiceless. 15.08% of the world’s languages haveejectiveplosives,13.08%haveejectiveaffricatesandonly2.22%haveejectivefricatives.Voiced implosives are produced on a combined airstream mechanism, i.e.laryngeal ingressive andpulmonic egressive.As far as the laryngeal ingressivecomponentisconcerned,thevocalfoldsareadductedandthelarynxasawholeispulleddownwards .Thiscreatesnegativepressureabovethevocalfoldsandair flows into the oral cavity to compensate the resulting pressure differencewithatmosphericpressure.Thevoicingarisesfromthefactthatthevocalfoldsare not firmly pressed together and as a result of the lowering larynx over a

staticcolumnofsubglottalair,asmallamountofairmayleakupwardcausingafewcyclesofvocalfoldvibration.Thesymbolsforvoicedimplosivesallhaveanupward-pointingrighthookextension.Voicedimplosivesoccurin11.31%oftheUPSIDlanguages:ofthe51languageswithvoicedimplosives,38arespokenontheAfrican continent.Curiously the IPAdoesnot include separate symbols forvoiceless implosives. Nevertheless, they do occur contrastively with voicedimplosives and voiced pulmonic plosives in languages such as Seereer-Siin inSenegal(McLaughlin2005),andLenduandNgiti intheDemocraticRepublicofCongo (Goyvaerts1988;Demolinetal.2002).Although in the spiritof the IPAprinciples, their phonemic contrastivenesswouldmerit separate symbols suchas[ƥ],[ƭ],[ƙ]and[ʠ],currentlythesesoundsarerepresentedaddingavoicelessdiacritictothecorrespondingsymbolsforvoicedimplosives:[ɓ̥],[ɗ̥],[ɠ" ]and[ʛ̥].Finally,clicksareproducedonalingualairstreammechanism:thetonguemakestwosimultaneouspointsofcontactwiththeroofofthemouth.Thebackofthetonguemakescontactwiththevelum,whilethebladeofthetonguetouchesthealveolar ridge, the postalveolar region or the hard palate. Between these twoocclusions,asmallamountofairistrapped.Subsequently,thebackofthetongueisdrawnbackwardwhilemaintainingcontactwith thevelum: this rarefies theair between the two closures. When the anterior closure is released, the airimplodes into the oral cavity to compensate the pressure difference. The non-linguisticuseofclicksiswidespread,butlinguisticallythesesoundsareconfinedto Khoisan and Bantu languages of the Southern African continent. The IPAsymbolsforclicksareunderspecifiedinthattheyrepresentthedifferentplacesof articulation only. This means that further specifications are necessary toindicatewhethera click isvoiceless, voiced,nasal etc.This canbeachievedbyaddingasuperscriptbeforetheclicksymbolasin[kǂ](voiceless),[gǂ](voiced)or[ŋǂ](nasal).3.2VOWELSYMBOLSINTHEIPAVowelsoundscanbedefinedassoundswithastrictureofopenapproximationwhichconstitutethenucleiofsyllables.TheIPAdistinguishes28vowelsymbolspresented on a trapezoid vowel chart which can be regarded as a schematicrepresentationofthevocaltract.TheIPAvowelchartisillustratedinFigure4:

Figure4:IllustrationoftheIPAvowelsymbolsplottedonavowelchartAll thesevowelsareoral inthat thevelumisraisedduringtheirproductionso

thattheairstreamescapesviatheoralcavityonly.Theverticaldimensionofthechart represents tongue height: vowels with a high tongue position (closevowels)arepositionedat the topof thechart,whilevowelswitha low tongueposition (open vowels) are placed at the bottom. In effect, this dimension issomewhatequivalenttothemannerofarticulationinconsonants.The horizontal dimension of the vowel chart represents the location of thehighestpointofthetongueinthevocaltract:justasintheconsonantchart,thevowelchartalsohasa left-to-rightorientation in termsofplaceofarticulation.Vowelsarticulatedatthefrontoftheoralcavityarelocatedleft,whilethebackvowels appear to the right with the central vowels featuring in between. InFigure4itcanbeseenthatmostofthevowelsymbolsareplottedinpairs:theleftsymbolofeachpairrepresentstheunroundedvowel,whiletherightsymbolrefers to the rounded vowel (i.e. produced with rounded lips) with the samedegreeofopeningandbackness.Thepreciseperceptualqualitiesof thesevowelsmayberatherdifficult todealwith by non-phoneticians. As a result, textbooks on phonetic notation oftenpresent descriptions of their perceptual qualities by referring to vowels inspecificlanguages.Thisgivesdescriptionssuchas[i]‘asinFr.si;Ger.wie.Withvalue more remote from cardinal in Eng. see’ or [ɛ] ‘as in Northern Eng.Pronunciationofpen,get;Fr.mettre(short),maître(long);Ger.Bett;Ital.pesca(peach),era(…)’(Principles,1948:8).Itshouldbepointedoutthoughthatsuchdescriptionsareonlytobeconsideredas rough approximations. More correctly, the qualities of at least 16 of thesevowelshavetobeappropriatelyregardedasCardinalVowels:theirqualitiesarenotbasedonthevowelsofanyexisting language.TheywerecreatedbyDanielJones as a set of fixed peripheral reference points with respect to which thevowels in all languagesof theworld canbedescribed. In creating these vowelqualities, Jones used well-defined articulatory and perceptual criteria, thedetaileddiscussionofwhich isbeyond the scopeof this article.The interestedreader is however referred to e.g. Abercrombie (1967) for an insightfuldescription of these design principles. Suffice it to say that their perceptualqualities can only be accurately illustrated and learnt by going back to theoriginalrecordingsofthesevowelsastheywerespokenbyDanielJoneshimself.Theserecordingsarenoweasilyaccessibleontheinternet.The number of qualitatively different oral vowels in languages of the worldvariesconsiderablyandrangesbetween3(inArabic)and24(in !Xu,Khoisan).The vast majority of languages have vowel systems with between 5 and 7vowels. The frequency of occurrence of vowels representing the variousarticulatorydimensionsdiscussedaboveisgiveninFigure5.

Figure5:Frequenciesofthedifferentvowelsinlanguagesoftheworldoutofatotal of 2,549 UPSID vowels. Underscored figures refer to rounded vowels(AdaptedfromCrystal1997:169)FromFigure5,itcanbeseenthatthevariousarticulatorydimensionsinvowelsarenotdistributedequally: front vowels tend tobeunrounded,while thevastmajority of back vowels are rounded. Rounding in the front vowels andunroundingofbackvowelsdoesoccur,butisrelativelyuncommoninlanguagesoftheworld.Furthermore,itcanbeseenthatlowvowelstendtobecentral.3.3EXTENDINGTHEROMAN-BASEDSYMBOLSTOCKFrom the discussion so far it seems clear that at least from an Englishperspective some of the symbols in Figure 1 are quite familiar from standardorthography: on the upper symbol row [p, b, t, d, c, k, g, q] areRoman letterswhicharefamiliartousersofRoman-basedorthographies.This stock of roman-based phonetic symbols has been extended by the IPA invariousways: besides significant import fromotherorthographies and symbolsets, several typographical changesandmodifications toexistingsymbolshavebeenused.Afirstextensionhasbeenachievedbyimportingsymbolsfromotheralphabets such as an older extendedRoman alphabet for English. Examples ofthis are the vowel symbols [æ] and [oe], which are still used today inScandinavianorthographiesandoccasionallyalsoappearinpresent-dayEnglishspelling (MacMahon 1996). Other symbols have been taken from the Greekalphabetsuchasa.o.[ɸ](Phi),[β](Beta),[θ](Theta),and[χ](Chi)torepresentseveralofthefricativesounds.A second extension has been the importation of symbols from other non-orthographic symbol sets. A good example of this is the symbol for the firstsound in the English word ‘shield’ which is represented as [ʃ]: this symbol isbetterknownastheintegralsymbolinmathematics.Inphoneticsitisreferredtoasan‘esh’.A thirdwayofobtainingadditionalphonetic symbols is the implementationoftypographical changes to existing symbols. A small number of symbols (i.e. 6)hasbeencreatedbysmall-capitalizing lowercaseLatin letters: [G], [N]and[R]

areusedtorepresentrespectivelythevoicedstop,nasalandtrillwithauvularplace of articulation, while [B] represents the bilabial trill. Furthermore, [L]referstothevoicedvelarlateralapproximantand[H]tothevoicelessepiglottalfricative.Somesymbolshavebeenarrivedatbyturninganexistingsymbolupsidedown:this is illustrated by the symbol for the uvular fricative [ʁ] and the voicelesslabial-velarfricative[ʍ].Other symbols have undergone more complicated changes such as mirror-imaginganexistingsymbolandthenturningitupsidedown,suchas[ɹ],[ə]and[ɐ].A final principle that has been a quite productive source of new symbols istypographicallymodifyingexistingromansymbols.Allthesymbolsreferringtoaretroflex‘placeofarticulation’havebeenobtainedbyaddingadownwardright-hookextensiontothecorrespondingsymbolforanalveolarplaceofarticulation:thustheset[ɖ],[ŋ],[ɽ],[ʐ],[ɻ]and[ɭ]isderivedfrom[d],[n],[r],[z],[ɹ]and[l].Something very similar has been achieved for the voiced implosive symbolswhichhavebeenderivedfromexistingsymbolsbyaddinganupwardright-hookextension,thuscreatingtheseries[ɓ],[ɗ],[ʄ],[ɠ]and[ʛ].It should be mentioned that such typographical modifications have importedsomedegreeoficonicityintheIPAalphabetinthatarecurringaspectofasetofsymbols represents a specificmeaning extending over awhole set of symbols.But whatever the modification or adaptation that has occurred, the IPA hasalwaystriedtoharmonisetheappearanceofthephoneticsymbolstoblendwellwiththeRoman-basedsymbolset,whichareallsetinRomantype,neverinitalictype.3.4DIACRITICSDiacriticsareextramarkingswhichcanbeassociatedwithmainsoundsymbolsto capture finer shades of pronunciation. The IPA alphabet contains 31 suchdiacriticsandtheirusageisillustratedinFigure6:

Figure6:IPAchartillustratingtheuseofdiacriticsItispossibletodistinguishseveralsubsetsofdiacritics.Themostextensivesetofdiacritics functions to describe modifications of the main sound segment ineither place or manner of articulation. Examples of place changes areadvancement [a�] or retraction [a̠] of place of articulation, while mannerchanges are exemplified by raising [e̝] or lowering [e̞]: in [e̝] the vowel ispronouncedwithahighertonguepositionthanforCardinalVowel[e],whilein[e̞]thetongueispositionedlower.A second set of diacritics enables a more detailed specification of phonationcharacteristics. These can be used to specify that a normally voiced speechsegment is devoiced such as in [n̥] or whether a normally voiceless soundappearsvoicedasin[tˬ].The third set relates to the release of consonants to capture processes likeaspirationofplosives[p],nasalrelease[d],lateralrelease[d]ortheabsenceofanyaudiblerelease[d].Thefourthsetallowsthetranscriptionofphenomenawhichnormallyarisefromco-articulation processes such as lip rounding, secondary articulations(labialization[b],palatalization[1],velarization[1]andpharyngealization[1]),nasalization[ɛ̃]andrhotacization[a˞].Finally,twofurtherdiacriticsspecifythesyllabicityofspeechsegments:theycanbeusedtoindicatewhetheranormallynon-syllabicspeechsegment(consonant)occurs syllabically [n̩] or that a normally syllabic sound (vowel) is used non-syllabically[e̯].

Diacriticsaregenerally,butnotalways,placedbelowthesymbols forthemainsound segments, butmay appear as superscripts positioned to the right of oraboveasymbol.However,itisperfectlyacceptabletoplaceadiacriticabovethesound symbol if this contributes to improving the readability of thetranscriptions. This applies particularly to sound symbols which significantlydescend below the writing line such as for instance the palatal [�] and velarnasal [ŋ]. In order to avoid clutter, the voicelessness of such segments can beindicatedbyplacing thediacriticabove themainsoundsymbolas in [��] and[ŋ̊].Ithastobekeptinmind,however,thatpotentialplacementconflictshavetobeavoided, especiallywhen transcribing nasalisation ([ẽ]) and creaky voice ([ḛ]),andbreathyvoice([e̤])andcentralisation([ë]).Intheseinstancesthepositionofthediacriticaboveorbelowthemainphoneticsymboldoesmatter.Inadditiontothediacriticsmentionedabove,thereareafurther9diacriticsandsymbolsrelatingtothesuprasegmentalaspectsofspeech.TheseareillustratedinFigure7:

Figure7:ListoftheIPAsuprasegmentaldiacriticsSomeofthesediacriticsrefertowordaccentplacement(primaryandsecondarystress) and the delineation of intonation groups,while others specify segmentlength.Finally, the IPA has 14 diacritics describing aspects of lexical tone in tonelanguages and pitch movements associated with word accents. These areillustratedinFigure8:

Figure8:ListoftheIPAtoneandword-accentdiacritics3.5IPAIDENTIFICATIONNUMBERSIt isuseful topointoutthateachIPAsymbolanddiacritichasbeenallocatedauniquethreedigitidentificationnumber.Thispermitsunambiguousreferencetothesymbolsine.g.emailandincommunicationwithpublishersofmanuscriptscontainingphoneticfonts.Thecodesforconsonantsbeginwith1(e.g.[p]=102)and the numbers referring to vowels begin with 3 (e.g. [i] = 301). Segmentaldiacritics begin with 4 (e.g. palatalisation = 421), while the suprasegmentaldiacriticsbeginwith5(e.g.longduration=503).Inthenumericalnotation,thediacriticsareassociatedwiththemainsoundsegmentsbymeansoftheadditionsign:e.g.[p]=102+421.4.EXTENSIONSOFTHEIPABesides the ‘traditional’ IPA symbols, this phonetic notation system has beengiven various extensions. The first important extension of the IPA wasintroduced in 1990 by the International Association of Clinical Linguists andPhoneticiansinClinicalPhoneticsandLinguistics(Duckworthetal.1990).Inthisextensionanumberofnewsymbolshavebeenintroducedtotranscribetypicalfeaturesindisorderedspeech.Anexampleofthisisthevelopharyngealfricative[fɳ] which is common in cleft palate speech or the bidental percussive [ʭ].Moreover, a significant set ofdiacriticshasbeenadded todescribe substantialdetail relating to various aspects of airstream mechanisms, phonation,articulation and timing. For example, the devoicing process of typically voicedsounds can be specified with great detail: partial devoicing [₍z̥₎], intial partialdevoicing [₍z̥], final partial devoicing [z̥₎].A striking featureofExtended IPA istheuseofbracketnotationssuchasparenthesesforsilentarticulations(speechwithout sound), double parentheses for unintelligible stretches of speech andcurlybracketsfordetailedprosodicnotation.ThesecondextensionisknownasSAMPAwhichisshortforSpeechAssessmentMethodsPhoneticAlphabet (Wells 1997). This is amachine-readable phoneticalphabetdevelopedinthelate1980sbyaninternationalgroupofphoneticians.Theaimofthisprojectwastoprovideastandardmachine-readableencodingofphonetic symbols.Theunderlyingprinciplewas to takeoverasmany symbolsfrom the IPA (such as ordinary p, b, t, d, k, g), but new symbols had to be

introducedforIPAsymbolswhicharenotmachinereadablesuchasIPA[ə]andIPA[ɑ]:thesearetranscribedinSAMPAas‘@’and‘A’respectively.WhereasSAMPAwasoriginallyrestrictedtoprovidemachinereadablesymbolsfor the sounds in a restricted number of European languages, X-SAMPA wasdeveloped by John Wells as an extension of SAMPA which covers the wholerangeofIPAsymbols.Inthissystemthevoicedbilabialimplosive[ɓ]forexampleis represented by ‘b_<‘ and the voiced glottal fricative [ɦ] by ‘h\’. Althoughunicode solves some of the problems that SAMPA and X-SAMPA aimed toaddress, both are still frequently used in computational phonetics and speechtechnologyapplications.Finally, itmayalsobeinterestingtonotethattheIPAalphabetisavailableinabraille version, the original of which was developed by Merrick and Potthoff(1932).Conceptually,inthisnotationsystemtheauralmediumisconvertedintoatactilemediumratherthanavisualone.TheBrailleIPAchartdoesnothavethefamiliar arrangement of the IPA symbols in terms of manner and place ofarticulationasintersectionsofcolumnsandrows:sincetheBraillesymbolsaresignificantly bigger than the print versions, the conventional arrangement ofphonetic symbols would yield a page which is too big for practical purposes(Wells-Jensen2005).5.LEARNINGTOUSETHEIPALearningtorecognisethesoundsoftheworld’slanguagesandmakingaccuratephonetictranscriptionsareessentialskillsforstudentsinmanylanguage-relateddisciplines. Nevertheless, acquiring these skills can be a challenging taskespecially in view of the increasing financial restrictions on academicinstitutions: traditional ear training classes are very labour intensive andconsequentlyexpensivetorun.This issolved inpartbyEartrainer(Verhoeven2003; Verhoeven&Davey 2007),which is a user-friendlymultimedia packagedeveloped atAntwerpUniversity andCity,University of London. This packageprovides a large stockof ear training exercises for students touseoutside theclassroom. On the basis of video recordings of an experienced phonetician’sdeliveryofawidevarietyofsoundsintheworld’slanguages,studentsengageindifferenttypesofeartrainingexerciseswithvariousdegreesofdifficulty.Intheeasier typesofexercises, studentshave toprovide thecorrect IPAsymbols fordescriptive phonetic labels or for individual sounds they hear. In the moredemanding exercises, learners have to complete partial transcriptions ofnonsensewords or they have to transcribewholewords. The interface of thelatterexercisetypeisillustratedinFigure9:

Figure9:InterfaceofEartrainerillustratingthetranscriptionmoduleIneachcase,Eartrainerprovidesinstantfeedbackonstudents’performanceandkeeps track of their progress. Furthermore, the transcriptions are analysed indetailsothatdirectedfeedbackcanbegivenonthespecificphoneticdimensionswhichindividualusersseemtostrugglewith.6.USINGTHEIPA:PHONETICTRANSCRIPTIONThetaskofphoneticallytranscribingaspeechsampleisapproachedsomewhatdifferentlybydifferentbranchesofLinguisticsandPhonetics.Incommontoallisthedesiretocommittopaperrathermoredetailaboutthespeechmaterialthancan be conveyed by orthographic transcription. Phonetic notation is used indifferent ways depending upon its intended purpose. Thus, a so-called ‘broadtranscription’ aims to capture the phonemic contrasts that are contained inspeech. For example, the word ‘water’ in standard British English may betranscribed as /ˈwɔtə/. The representation of the basic sound contrasts in theword is clearwhen comparedwith the transcription of /ˈwɔmə/(‘warmer’) or/ˈdɔtə/(‘daughter’), inwhichthereplacementofasinglecontrastivesound(or‘phoneme’)would result in a change inmeaning. It should be noted that suchbroad or ‘phonemic’ transcription is usually presented between obliques. Thismaybecontrastedwith ‘narrowphonetic’transcriptioninsquarebracketsthataims to includedetailsaboveandbeyond thatwhich is required toconvey themeaningsofthewords.In theword ‘water’ e.g., itmaybe relevantperhaps to indicate that themedialconsonant is realised as a glottal stop [ˈwɔʔə] (asmight be found in a Londonspeaker),orthattheinitialsoundisproducedasavoicelesslabialvelarfricative

i.e.[ˈʍotə],orperhapsthatthesamplederivesfromarhoticspeakerwhousesanr-colouredfinalschwaandamedialtap[ɾ]ratherthanaplosive[ˈwɔɾɚ].Insuch‘narrow’ phonetic transcription, the fine detail of pronunciation is includedwhereasinthephonemictranscriptionitisimplicitlyacknowledgedthattheuseof [t] or [ʔ] does not change the intended meaning of the word. In short,therefore,theleveloftranscriptionisselecteddependinguponthetaskinhand–andineithereventtheIPAwillprovidethespeechscientistwiththenecessarytranscriptiontools.It may be mentioned in passing that producing a transcription may be apainstakinglydetailedjob.Onedifficultyisthatthesoundmediumcarryingthespeech is intrinsically transitory.Toovercome this ephemeral characteristic ofspeech, the phonetic field worker will frequently make a sound recording,nowadaysprobablyusing solid state technology.When the speech is replayed,the sound of course remains ephemeral but at least it may be repeatedlyreproduced,refreshingtheimpressionisthelistener’sauditorycortex.Whether‘broad’or‘narrow’,thetranscriptionproducedinthiswayissubjective,basedasitisupontheinterpretation,albeitskilledone,oftheperceiver’sear.Inan attempt to make the process more objective, phoneticians (qua speechscientists)willtransferthespeechrecordingtoacomputer,andexaminesoundspectrograms of the sound in order to test specific auditory impressions. Toreturn to our example one last time, some corroboration of a rhoticpronunciation of theword ‘water’might be empirical evidence of a somewhatlowered third resonance (or formant) in the final syllable. By means of suchauditoryandacousticphoneticmethods,therichdetailofspeechmaybewrittendownandthoroughlyunderpinnedbyscientificmethods.Notes1. It should be noted that vertically arranged analphabetic systems have beendevelopedaswell.AgoodexampleofthisisthatofThomasWrightHill(1860).Such vertical arrangement seems evenmore impractical from a typographicalpointofview.2. The frequency calculations of sound categories reported in this paper arebasedonthesoundinventoriesof451 languages,317ofwhichwereoriginallypublished in Maddieson (1984). User-friendly access to this database is nowavailableonhttp://web.phonetik.uni-frankfurt.de/upsid_find.html.References

Abercrombie,D.1967 Elementsofgeneralphonetics.EdinburghUniversityPress.Bell,A.M.1867 Visible speech: The science of universal alphabetics; or self-interpretingphysiological letters, for the writing of all languages in one alphabet. Simpkin,Marshall&Co;N.Trübner&Co.Catford,I.1977 Fundamentalproblemsinphonetics.EdinburghUniversityPress.

Crystal,D.1997 TheCambridgeencyclopediaoflanguage.CambridgeUniversityPress.Demolin,D.,H.Ngonga-Ke-Mbembe&A.Soquet2002 Phonetic characteristics of an unexploded palatal implosive in Hendo.JournaloftheInternationalPhoneticAssociation32:1–15.Duckworth,M.,G.Allen,W.Hardcastle&M.J.Ball1990 ExtensionstotheInternationalPhoneticAlphabetforthetranscriptionofatypicalspeech.ClinicalLinguisticsandPhonetics4:273–80.Goyvaerts,D.1988 Glottalizedconsonants:anewdimension.BelgianJournalofLinguistics3:97–102.Jespersen,O.1889 The articulations of speech sounds represented bymeans of analphabeticsymbols.N.G.Elwert.Laver,J.1994 Principlesofphonetics.CambridgeUniversityPress.Maddieson,I.1984 Patternsofsounds.CambridgeUniversityPress.Macmahon,M.K.1996 Phonetic notation. In P.T. Daniels&W. Bright (eds.)Theworld’swritingsystems:821–46.OxfordUniversityPress.Merrick,P.W.&W.Potthoff1932 A Braille notation of the International Phonetic Alphabet. The RoyalNationalInstitutefortheBlind.Mclaughlin,F.2005 Voiceless implosives inSeereer-Siin. JournaloftheInternationalPhoneticAssociation35:201–14.Pike,K.L.1943 Phonetics: A critical analysis of phonetic theory and a technic for thepracticaldescriptionofsounds.UniversityofMichiganPress.Principles1948 TheprinciplesoftheInternationalPhoneticAssociationbeingadescriptionof the InternationalPhoneticAlphabetand themannerof using it, illustratedbytextsin51languages.UniversityCollegeLondon.Verhoeven,J.2003 Ear training on line as an educational tool for the teaching of IPAtranscription. InL.Polka,M.Sundara,M.J. Sole,D.Recasens& J.Romero (eds.)Proceedingsofthe15thInternationalCongressofPhoneticSciences:3085–87.Verhoeven,J.&R.Davey2007 A multimedia approach to eartraining and IPA transcription. In J.Maidment (ed.)Proceedingsof thePhoneticsTeachingandLearningConference:1–4.Wells,J.C.1997 SAMPAcomputerreadablephoneticalphabet.InD.GibbonR.Moore&R.Winski(eds)HandbookofStandardsandResourcesforSpokenLanguageSystems:PartIV,sectionB.MoutondeGruyter.Wells-Jensen,S.2005 TheBraille InternationalPhoneticAlphabetandotheroptions: theblindstudent in the phonetics classroom. Journal of the International Phonetic

Association:35:221–30.WrightHill,T.1860 Onthearticulationofspeech.