Hindi PB List for Speech Audiometry and Discrimination Test N.S. De Bangalore

Speech audiometry is now an accepted essential test in the functional evaluation of auditory apparatus. Development of modern ear surgery and sophisticated hearing aids have enhanced the need for such quantitative speech tests. In western countries, considerable work has been done towards standardisation of these tests. Due to linguistic barr, iers, preparation of an internationally acceptable test material has not been possible. In India, multiplicity of languages and varieties of dialects have offered a great obstacle in the evolution of common speech material for audiometric tests. However, working in the Armed Forces, where one comes across a cross section of personnel from almost all parts of India, one is struck by tho fact that even though the mother-tongue of the various individuals varies a great deal, their grasp and intelligibility of the commonly spoken Hindi is quite adequate for discrimination tests of hearing. To test this fact scientifically, a comparative study was made with English, Hindi, Bengali and nonsense speech material and the conclusion of this study are as under :-Sqn Ldr. N.S. DE, ENT Specialist,Institute of Aviation Medicine,Bangalore.

(a) Persons consistantly produce better and optimum discrimination score in their mother tongue as compared to other languages. (b) Discrimination score obtained with material in second language/languages is fairly reliable.

(e) Subjects having

a working knowledge in two languages, none of which are their mother tongue, discriminate better in a language of Indian origin than in a language which is foreign in origin.

(d) Hindi, Bengali and Indian English to a person of Indian origin whose mother tongue is none of these languages, is not unfamiliar in the same sense as American English or any other foreign languages, say Spanish or Arabic. (e) Persons do not discriminate well when tested with nonsense syllables. It is therefore felt that speech material composed of the commonly spoken Hindi words can be used in most parts of the country and a study of such a material is undertaken in this article. 64

Speech Audiometry Material Various types of material have been tried for speech audiometry, eg. numbers, nonsense syllables, mono and polysyllabic words and sentences. The best known lists of words for discrimination testing are probably the PB-50 word lists. These are lists, 50 words each of reasonably familiar monosyllables and were developed at the psychoacoustic laboratory at Harvard. These lists meet the following criteria :--

discouraged a concentrated effort in, preparing speech test material in any one language.

Hindi Speech Audiometry Material Gupta, OP, Misra RN et al (1960) compiled a list of spordiac words in Hindi. In the same work they also carried out a phonetic analysis of Hindi language.-

Hindi as a language is the most widely used of the languages of the South Asian sub-continent. It is spoken in the greater part of North (a) Monosyllabic structure India, in states of Delhi, Uttar (b) Equal average difficulty Pradesh, Rajasthan, Madhya Pradesh, (c) Equal range of difficulty Bihar and parts of Punjab and (d) Equal phonetic composition Haryana, as well as in large north and (e) Composition representative of central Indian cities. Statistics show American speech. that between 250 to 3t70 million people (f) Words in common usage. speak in Hindi, thus making Hindi one of the major languages of the The American PB series consists of world. Besides its being 'Rastrabhasha" 20 lists, each containing 50 monosyll- all Armed Forces personnel have t ables. They cover a wide range of know Hindi for their day to day work difficulty to make them adequate for most type of articulation comparisons. Hindi is a phonetic language, The spread of difficulty is approxi- unlike English which is highly nonmately same in each test and each test phonetic. Hindi is commonly written has nearly the same average difficulty. in Devnagri script which has been Furthermore, the lists have very nearly evolved on scientific basis so that the the same phonetic composition quite scripts correspond to and are actually similar to that of English language. synonymous with the phonemes. Rare and unfamiliar words have been omitted as much as possible and very Phonetic Character of Hindi Language few of the words are extremely easy or extremely difficult under conditions in In this study, phonetic structure of which the whole group of lists would Hindi Language has been taken as it obtain an average articulation score of occurs in Armed Forces personnel, about 50 percent. who represent a fair cross section of the Indians of all states. It has been Speech audiometry based on the seen that 10 basic vowel sounds with American experience has been adopted their nasalised and aspirated forms in a dozen other countries; notable along with 31 consonantal phonemes among them are England, France, were quite adequate to transcribe all Germany, Italy, Yugoslavia, Holland, the Hindi words. Finland, Sweden, Spain, Denmark, Russia and Japan. Advantages with Phonemes used in this work workers in all these countries have been that most of these countries are A. VOWELS unilingual and development of one uniform speech test material was easy, The ten basic vowel sounds are a~ whereas multilingual character of India follows :-65

Ind. 3". Otol. Vol. X X V , No. 2, June 1973

Localisalion of Hindi Speech Sounds in Frequency Spectrum Nasalisation and Aspiration

Filter was used to determine the formants of Hindi vowel and consonant sounds in the frequency spectrum. Considarable narrow frequency bands were obtained by using both high and low cut-offcomrols o f the filter. By pronouncing the phonemes through the micro.circuit of the Allisons 4C Audiometer, the sounds were plotted on graphs. Figure 1 shows the relative position of the formant areas schematically. The understanding of the formant areas are important to have an idea about the ditticulties a person may have with a language when he loses function of a part of the frequency range. The knowledge of the formant areas are essential to formulate planned

All these ten basic vowels occur in nasalised and aspirated form in the language B.

Consonants

PLOSIVES

~

NASALS SEMIVOWELS FRICATIVES FLAPPED ASPIRATED

~r

~T

~r~T

gh~C~WSNcu 12jr

:~5o

5oo

/~V ~ P S

7~'o

r t

I

..........

~o

:$

1t"

:US P 9

.,4

q::

u~

qo

Fig. I The relative position of the formant areas of Hindi speech sounds.

Hindi PB List/De

66

auditory training for deaf and dumb children with high tone loss or islanded audiogram (See Figure 1). The relative intensities of Hindi speech sounds were worked out with the vu matter and micro circuit of the audiometer. The consonants could not be pronounced independently, minus the post-consonantal (Aw) which to some extent has modified the figures shown in the diagram. All the same, it gives an idea as to the relative intensities of different speech sounds in Hindi and the approximate range of intensity required to make them audible as a whole or in combination (See figure 2). Phonetic Analysis of Hindi Language

In this work, Hindi language material was taken from different sources so as to obtain, as far as possible, a fair representation of the language as a whole. Language material almost in equal proportions were taken from Hindi primers, Hindi daily newspapers, day-to day conversational Hindi in the Armed Forces and commonly occuring Hindi words from Hindi dictionary. The material was then transcribed phonetically by using a method o f broad transcription. Following this procedure, it became

easier to pick up the statistical figures on following lines : ~ (a) Occurance of vowels. (b) Occurance of consonants as initial consonants. (c) Syllabic structure of the language. The figures thus obtained were utilised for preparing the PB list of words. (See tables 1, 2 and 3) TABLE I Percentage of vowel sounds separately in Hindi language ~r

--

23.27%

~

~

24.63%

~

-~ ~

~ ~r

7.86 % 8.27% 4.40%

B. Nasalised Forms

9" ~t 5"

---

1.47% 0.95%

-

0.50%

---

0.23% 0.16%

~" ~"

~( -0.04% Total 100% of Vowels only.

Fig. 2 The relative imensitir of Hindi speech sounds

Ind. Y. Otol. Vol. XXF. No. 2, June 1973

0.54% 1.10% 0.03%

~ ~'T

C. Aspirated Form

4q

67

~ ~'~ ~'~

1.93% 11.65% ,~.26 70 5.42% 1.18o,/0

1.72% 0,78%

T A B L E II Percentage of Consonants taken as initial consonants in Hindi language, 5.4% 2.5% 4.0% 3.0% 3.5% 2.0% 4.0% 2.0% 2.5% 1.4%

~r

~, a e

~.

~ ~ ~ ~ ~ ~ ~ ~ ~r ~

7.5% 1.4% 7.8% 4.0% 5.5% 4.0% 1.4% 4.0% 3.870 3.0%

1.5%

~

7.8%

1.4% 4.0% 1.5% 4.0% 1.5%

~r ~. ~ ~,

2.7% 0.0% 0.0% 3.5%

TABLE III Syllable structure of Hindi Language CVC

--

45%

cv vc

--

30% 20%

Others

---

labic words are to be taken. Once this is done the words are to be arranged in such a way that a 50 word list will contain 50 vowels in the same proportion as has been obtained in the phonetic analysis. The p r o p o r t i e n of initial consonant are now to be checked and suitable wo~ds are to be substituted so as to make the percentage of initial consonants in the list approximately conform to the percentage obtained in statistical analysis. The words in the list are now to be re-arranged in such a way that the list becomes a homogenous one i.e. similar sounding words are suitably dispersed. To give an example as to how tc~ proceed with the work of preparing a list of say, 10 words when the required statistics are available the following can be taken as a sample. Available statistics (A) Vowels a--50%~5a e--20% =2e 0--20% =20 u - - I 0 % = lu

5%

Definition of Phonetically Balanced (PB) word List PB list of words consist o f a group of single syllable words so selected that the frequency o f occurance of speech sounds within the group is the same as the frequency o f occurance of the same sounds in an average vocabulary of conversational language material. PB words are used to determine discrimination scores for speech. Procedure to be adopted for making PB List of words As per the requirements a PB list o f words should consist of (a) m o s t familiar monosyllabic words (b) should have a representation of phonemes as per the occurance of phonemes in the language. To start with all commonly occuring and familiar meaningful monosyl-

Hindi PB List/De

(B) Initial consonants b - - 2 0 % ~ 2b k--10%=lk (as in cap) f--10% ~__lf m - - 20 % ~ 2 m p--10%=lp r--20%--2r s--10~ ls (C) Commonly occuring monosyllabic words. Bad, moon, tea, cap, ten, seed, meat, fan, doll, sun, pen, cat, fish, rat, bell, root. With all the above figures in hand, the list can be compiled as follows : - Bad Bell Man Root Sun

moon pen Cap Fan Rat 68

As mentioned earlier it is impossible to balance a list as regards the occurance of final consonants due to the pre-imposed statistics of initial consonants and vowels along with the other requirements that the words should be monosyllabic, familiar and meaningful. Preparation of the Lists

To begin with, all the cvc, vc and cv types of monosyllabic words were collected and compiled into a list. F r o m this list words that were unfamiliar or difficult were discarded. The remaining words were used to prepare eighteen lists o f 50 words each. All these lists were phonetically balanced. The lists were tested on normal Hindispeaking subjects. The words that were constantly missed were replaced by suitable easier words. After initial trials a final set of six 50 word PB lists of Hindi words were accepted for clinical trial and use. The lists are reproduced at the end of this article. Standardisation of the test material

material thus prepared was obtained by carrying out clinical tests with these lists. A few minor alterations were done as a result of these trials so as to obtain a fair degree of uniformity amongst the lists. In the final shape the lists gave uniform articulation curve and consistantly similar discrimination scores (See Figure 3 and Table 4). TABLE IV

No. o f Ears Tested

%Discrimination Average Score at 40 dB discriminaRe. 0.0002 tion for this Dynes/Sqn. Cm Group

1

2 3 4 5 6 7 8 9 10

96 96 100 98 100 96 100 98 100 100

Further standardisation of the test ~PEEC/~

RUDtOd:RRM OOrE

/

tl

[

/o

Fig. 3

~

/ 8o

[Average normal articulation curve with Hindi PB lists

69 Ind. J. Otol. Vol. XXV, No. 2, June 1973

98.4%

Instruments and Testing Conditions The audiometer used was Allisons 4C type audiometer. The tape deck was used to record and reproduce the word lists. Recordings were done at night to obtain reasonably quiet conditions. An experienced male person was used as a talker. 'Boliye' was used as a warning phrase before pronouncing each word of the lists. The room where all these tests were carried out is sound treated with ambient noise of about 30 db at day time. Individuals were tested through earphones ; they were asked to repeat the words as they perceive them. The score was kept by the operator and the same expressed as percentage o f words heard correctly or percent discrimination score. Variables Operating in a Controlled Speech Test There are many factors which influence discrimination score and the shape of articulation curve. Understanding of these factors can give us a better idea towards the clinical evaluation of certain discrimination score obtained. The variables are : - -

A--Type of Test Material The composition of test list is important. Lists that are too easy or lists which provide too many clues to the listener tend to under estimate the disability. Conversely lists that are too difficult or full of unfamiliar words tend to produce unreliable results,

B - T h e Talker Some persons talk very clearly and are quite easily understood. There are others who speak with pronounced dialect or foreign accent and are less easily understood. To obviate these anomalies, it is advised to use practiced speakers.

C--Recording and reproducing technique To avoid falacies, attention should be paid to the following points :

ttindi PB. Lists[De.

(a) Speed of recording and speed of reproducing should be the same. all back(b) While recording, ground noise should be eliminated.

(c)

While recording as well as reproducing the amplification should be checked correct in the vu meter.

D--The Operator The operator while monitoring the subjects responses may record a correct response as a wrong one and vice versa. This can happen even with experienced operators.

E--The Listener The subject's full cooperation at the time of testifag is essential for obtaining a reliable figure o f his discrimination score. People who are keen to hear, discriminate better than people who are not interested in the test.

F--Level of Presentation In routine investigations, Hindi PB lists should be presented to the subject earphones at an intensity level of 35 dB above pure tone threshold to obtain the optimum discrimination score. Where marked deafness is present, it is advisable to plot a detailed speech curve by presenting a numbet of pB word lists. Results of Clinical Trials Carried out with Hindi PB Lists

A--Normal persons (a) All normal persons whoso mother tongue is Hindi irrespective of education and service background produced a discrimination score o f 98100% at M.C.L. (See Fig 4). 70

l!I I{

~TG --

f

I

iI

ii

_

P8 M,~x Peo _ _

_

Io

ro

i i

~l

-

-

- -

so

30 5 o o lV D

6o

70 ~ /Ar "4

PA~$$uRs

?9

seo

Fig. 4 Average normal speech audiogram of normal Hindi speaking persons.

--

~

I

- -

--

L

/

II

,

I

__

l.

.

.

.

P8

:

ec~rq~5

P,q~#$trR~ I/V" b ~

Average normal speech audiogram of persons whose mother tongue is a language other than Hindi, but who had some knowledge of Hindi.

(b) All normal persons whose mother tongue is a language other than Hindi but who had some knowledge of Hindi language producedan average score of 92-96% at M.C.L. (See Fig. 5).

B--Persons with conductive type of deafness (a) Discrimination score in pure conductive deaf patients were the same 71

t

S $OONa

Fig. 5

- -

as normal persons under similar circumstances. The M.C.L. in these cases were considerably higher (See Fig. 6) (b) Early otosclerotics discriminate like normal persons. Advanced Otosclerotics show a tendency to produce discrimination, loss depending on the degree of deafness (See Fig 7 &8)

Ind. J. Otol. Vol. XXV, No. 2, June, 1973

~.

AvM~,.R4N

$PEs

/

m

~

4r4ml N 8

PB M ~

eg

g r

Fig. 6

_

~ I

Speech audiogram of a conductivedeaf patient

.

i

ii

:

I

--

--

II

i

I

L ~

Fig. 7 Speechaudiogram of early otoscleroticdeafness

C--Persons with perceptive type of Deafness Discrimination tests are valuable for assessment of cases of perceptive deafness. It gives a better idea of the functioning of the hearing apparatus and extent of disability (See Fig 9 (a) & 9 (b))

Itindi PB List~De

Reliability of Hindi PB Lists in a Linguistically Mixed Population Clinical trials were carried out on persons belonging to different linguistic groups serving in the Armed Forces of India. Though differing in their mother tongues all of them had a working knowledge of Hindi. The 72

SP6G~.t~ R~t~toaaR~f

~r~

~.~

/4

o+

m~x 7~

/ i

[]

~

_

-

Fig. 8 Speech audiogram of advanced otosclerotic patients

N 4..t,I,E ~ , ~ .

-

-

-

- -

aMuAt+

i.

....

+ - - r

pRE,~.$uR~C tt~ # B

Fig. 9 (a) & (b) Speechaudiogram of perceptive deafness

~1,,+ 11.r "

L

~

=I

-

I

II

. / o

ro

zo

.30

~o

$ o u ~'+

6o

~

7o

PRE $5 u R E

+po av"

D8

?,o-

too

results of speech tests did not show any marked difference between persons whose mother tongue was Hindi a n d those others to whom Hindi was a second language. It is felt t h a t a simple f o r m of Hindi, as spoken by the common man, can be used as a suitable material for PB word lists in speech audiometry throughout the country. Hindi PB Lists Versus Nonsense Syllables To explore the possibility of making an universally acceptable speech test material for Indian patients, trial was carried out with PB list o f nonsense syllables. The discrimination scores obtained were widely varying, inconsistant and unreliable. Hence use of PB nonsense syllables are n o t to be recommended for standard speech tests.

zt~r zu

qT~T

~'T~-~ ~TT~

List 3 ~l'T~

q~T

~T:q" ~T~

tq"T

~TIT

N'g[

~T~T

~r-g

~I'TTT

z~rq m~

~'T~ ~Ttl" ~;'~.

~'T11"

~T~

~r

~"r ~TTTT

qzr

g't~ ~z

List 4 qr

Hindi PB List I

f~r~

~T~"

['tq gin

f~

1:1~

~TX

,t:F;~

~Tg

~iT'(

r ~I"T1:T

rTI~T

~T

slTq

e~r~r

3,:r

:L':TTZl"

,fr

ZI~T

~r

NTN" List 5

f~T~r List 2

;TT~

gig ~3"g ~'Tq'

~:q"

~T~T

~[~

~T

q~r

~'~

~r~

~TTF ~lq"

~TTIT

~"r~

~T~T

gig

f~z

Z~rt qT~

~(Tg

~T~

Hiudi PB List/De

~r

~T~F 74

List 6

Acknowledgement

fmr

~r

~rm"

~r

~rrw

~r~r

~rr~r

~z

q.

~z

I am grateful to Colonel Daryao Singh, Senior Adviser in Otorhinolaryngology for the Armed Forces under whose guidance and supervision this work was conducted in the ENT Department of Army Hospital Delhi Cantt. REFERENCES

~t~

~r

tr~r

~;'1~

Summary The purpose of the present work was to compile and standardise test material for speech audiometry in India. Phonetically balanced list of words in Hindi has been prepared and used for discrimination tests. Six PB Lists of words have been prepared. Instruments, testing conditions and different variables operating in a controlled speech test and their effects on the final discrimination score has been discussed. The clinical application and method of presentation of the lists to the patients have been discussed, It is felt that in a multilingual country like India, Hindi speech material can be accepted for routine speech audiometry for all persons who have even minimum knowledge of Hindi.

75

1.

Carhart, R. : Basic Principles of Speech Audiometry Acta Otolaryngologica, 40 : 62, 1951.

2.

Carhart, R. : Problems in measurement of speech discrimination, Arch. Otolaryngol, 82, 253, 1965.

trr~r

3. Cooper RAF and

Merchant, N . R ! Speech Audiometry. Indian Journal of Otolaryngology, 10 : 193, 1953.

4. Egan JP. : Articulation testing methods. Laryngoscope, 58. 955, 1948.

5. Fry D.B., and Kerridge, P.M.T. : Tests for the hearing of speech by deaf people. LANCET, 236: 106, 1939. D.B., : Word and Sentence test for 4. Fry use in speech audiometry: Lancet, 2,/7]95: 197, 1961.

7. Gunnar Liden, : Acta Otolaryngologica, supp 114, (1954).

O.P. Misra, R.N., and Bhatia, 8" Gupta, M.L. : The significance of phonetics of speech to the otolaryngologist ; Indian Journal of Otolayngology, 14 : 1 (1962).

Ind. J. Otol. Vol. XXV, No. 2, June, 1973