The Journal Of Laryngology And Otology - Cambridge

Testing Hearing Efficiency of Aviation Candidates theoretically, be an easier task to measure the difference between individuals whilst keeping the listening conditions uniform. a. TYPES OF TEST ia. Sentence articulation tests. A sentence articulation test is a test designed to measure the percentage of sentences correctly received by the listener. Most workers in this field amend the definition of sentence articulation and say " the percentage of sentences of which the sense is correctly received ", thus allowing for minor verbal errors which do not materially alter the sense. While this seems to be the most sensible way to regard sentence articulation, it does bring in the possibility of variations in the judgment of the people marking the responses to the test, since it is not always easy to agree when the sense of the sentence has been materially altered. However, it is possible to get over this difficulty by somewhat modifying the technique of marking (see p. 161). A number of different types of sentence articulation test have previously been used. The following are the most common : i. Question and answer. A series of simple questions is sent to the listener who is deemed to have received the sense of the question correctly if he gives what appears to be a sensible reply. ii. Simple sentences. A large number of short sentences is read to the listener who records his reception of them either by writing them down or by speaking them back to the speaker, who checks the accuracy of the reception. Articulation tests usually have to be carried out in difficult acoustic conditions and if the second method of recording is used it is often necessary to provide an independent high-fidelity channel to carry the listener's reply. In any case, the method of writing down the sentence is certainly safer if also rather more laborious. When the present research was begun, it was accepted as axiomatic that a measure of a listener's sentence articulation over a given system in given conditions was the nearest one could get to a direct measure of his ability to use that system for listening to current speech. It was further adopted as a basic principle that a reasonably close correlation must be proved between the results of sentence articulation tests and those of any other form of test which might be used before the validity of the latter could be accepted. 2a. Word articulation tests. A word articulation test is one designed to give the percentage of words correctly received by the listener. It is commonly understood that in this type of test the words should be sent either without linguistic context or in a neutral context, i.e. they should be sent as isolated words or in a carrier sentence which gives no clue to the meaning of any test word. The selecting of words for these tests is not an easy matter. Theoretically, the frequency of occurrence of words in the test should 143 https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

Dickson, Simpson, Fry, Swindell, Brown reflect their frequency of occurrence in the kind of speech which is to be sent over the system. To devise a test conforming to this principle would necessitate the long and laborious task of making word counts for various types of material. If it is assumed that the system will be used for all kinds of speech material, then it is possible to use some standard word count such as that made by Dewey* for American English. Unfortunately, no such count exists for British English. An alternative method is to base the selection of words on the frequency of occurrence of sounds in current speech. It can be argued that words are recognized chiefly by means of the recognition of speech sounds and if therefore we make a selection of words which are reasonably familar and in which the distribution of sounds corresponds to that of current speech, we shall have a test which is likely to give a reliable measure of word articulation. Even so, a further complication is added by the importance of rhythm for the recognition of words. This factor can be ruled out by using monosyllables exclusively but a test of this character is of doubtful validity as a means of measuring word articulation since it takes no account of the occurrence of different rhythmic patterns. 3a. Syllable articulation tests Most so-called syllable articulation tests are in reality sound articulation tests (see section 3 : 4a), that is to say they bear very little relation to " syllables " in the linguistic sense of the term and are really a particular form of sound articulation test. If we take a passage of connected speech and examine its component syllables we find that they differ not only in their sounds but also in the stress and pitch which are used in pronouncing them. These last two factors have a great effect on the recognition of the syllables and no test could be considered a true syllable articulation test which did not allow for these factors. No attempt has ever been made to devise such a test and indeed it is very doubtful whether such an undertaking would ever be worth while. Probably the results of such a test would add very little to the information which can be obtained more easily by means of sound, word and sentence tests. The " syllable " articulation tests in present use consist of arrangements of speech sounds to form meaningless syllables ; these may be of the form consonant-vowel-consonant (CVC), vowel-consonant (VC), or consonant-vowel (CV). The frequency of occurrence of sounds in such a test may or may not be weighted in accordance with the frequency of occurrence of the sounds in current speech. The test is marked by marking any one syllable right or wrong. Experience in various laboratories has shown that the scores for this type of test tend to be lower than those for a sound articulation test possibly because in one test the probability of getting a complete syllable i.e. a particular group of * Dewey, C. " Relative frequency of English speech sounds " (Harvard University Press, 1923). 144 https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

A NEW METHOD OF TESTING THE HEARING EFFICIENCY OF AVIATION CANDIDATES.— K. D. I). DICKSON, J. F. SIMPSON, I). B. FRY, G. E. SWINDELL, R. E. C. BROWN. FIG. I. General view of rack-mounted equipment. [face p. 144 https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

FIG. 2. Close-up view of scanning unit. https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

Testing Hearing Efficiency of Aviation Candidates three consecutive sounds right is less than that of getting the same three sounds right when no restriction is placed on their positions. 4a. Sound articulation tests Various tests have been devised to measure the percentage of speech sounds correctly received over a system. Some of these employ meaningless arrangements of sounds and others use meaningful words to test the reception of particular speech sounds. It can be stated in general that the tests using meaningless arrangements are more reliable as a means of measuring sound articulation, provided they are used with trained crews. With this type of test the listener has to rely entirely on his hearing of the sounds transmitted ; in a test in which meaningful words are used, the sounds are attached to the linguistic framework of the listener's language and the personal differences between listeners, differences for example in their vocabulary or in their quickness of wit, become more important and have a greater effect on the results of the test. Whichever type of test is used it is advisable to make allowance for the frequency of occurrence of the sounds in current speech. This may be done in the selection of the sounds to make up the test or it may be done by weighting the results. The latter involves a good deal of laborious work on all the results obtained by the test and it is usually worth while to do the weighting when making up the test material. When the necessary sounds have been selected in the right proportions, they can be combined to form either meaningless or meaningful syllables. In both cases the responses to the test are marked to give credit for every sound correctly received. A very good type of sound articulation test using meaningful words has been devised by Harvey Fletcher and is given in his " Speech and Hearing " (1929) p. 263. In this test, monosyllables of the types CVC and CV are used but in each syllable only one sound, either vowel or consonant, is marked. For example, in a consonant articulation list there are a number of words all beginning with the combination wi-*. A final consonant is then added to form words such as wit, win, etc., and the final sound is the only one to be marked in these words. Other series of words test the reception of initial consonants and of vowels (see Appendix II, p. 197). The different types of articulation test mentioned in the preceding paragraphs were all considered from the point of view of their suitability for use as a hearing test. It has already been said that the sentence articulation test was held to be the most direct measure of the aptitude it was required to assess. Practical difficulties however, make the sentence test impossible for adoption as a standard hearing test. If the test were to give reliable results it could not last less than 20-25 minutes and this would be too long a hearing test to use for the examination of some thousands of candidates. Consequently, it was necessary *See Appendix I for an explanation of the phonetic symbols used in this report. 145 10c https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

Dickson, Simpson, Fry, Swindell, Brown to examine the other types of articulation tests to find out which was the most suitable, bearing in mind always that it was necessary to demonstrate a significant correlation between the results of a selected test and those of sentence articulation tests. One type of test had to be ruled out of consideration, the type of sound articulation test which employs meaningless syllables. This kind of test can only be used with trained listening crews who have been taught some systematic method of writing down the sounds which they hear. In a hearing test naturally there can be no question of practice for the candidate and the result will be expected from his first attempt. The types of test left for use in the experimental work which is now to be described are therefore sentence articulation tests as the direct measure of listening aptitude and forms of word and sound articulation tests with meaningful syllables as possible alternatives. 4. Equipment for Experimental Tests In this section a general description will be given of the apparatus used in the experimental investigations leading up to the development of a new and more satisfactory hearing test. It was decided at the outset that in order to maintain standardized testing conditions all articulation tests should be recorded on gramophone discs and reproduced electrically. Most of the electrical equipment required was mounted on the standard four bay Post Office rack shown in the accompanying photograph, Fig. i. Wherever possible the input and output impedances of each unit were fixed at 600 ohms, so that they could be connected together directly to form any desired channel. All connections were made on a jack panel using short screened leads. This arrangement was found to be quite satisfactory for most purposes, the small amount of hum induced in the lines being insufficient to interfere with our tests. a. PROVISION OF THE NOISE BACKGROUND For subsequent experiments we require to be able to produce a noise field similar in frequency and overall level to that existing in a typical modern Service aircraft. Two methods for producing such noise suggest themselves. In the first the actual noise in the fuselage of an aircraft in flight is recorded and reproduced in the testing room, frequency correction being applied to compensate for any loss during recording. In the second method the noise is produced synthetically by amplifying suitably modulated electrical oscillations. The second method has certain advantages, notably ease of production and control, but it is possible that the resultant noise, though satisfactory from the point of view of level and characteristic, may not sound like an aeroplane engine. It was felt that a noise which really gives the impression of an aeroplane in flight would be preferable to one which does not sound so realistic. Another argument in favour of using the recorded 146 https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

Testing Hearing Efficiency of Aviation Candidates noise is that there is in existence a standard Service " noise generator " which is used for providing background noise for various types of synthetic training. There did not seem to be any reason why this generator, which makes use of an engine noise recording, should not be satisfactory for our purpose, and it was therefore decided to use it as the basis of our standard aero-engine noise. The steps in the development of the final satisfactory noise will be described in some detail in order to give an idea of the difficulties which were encountered in obtaining an adequate reproduction of aero-engine noise. The first requirement was a testing booth with sound absorbing walls which would confine the noise. The internal dimensions of the booth which was constructed for us were : height 9 feet, length 5 feet, breadth 4 feet. No elaborate sound proofing was attempted as it was considered sufficient to reduce the noise outside the booth to a level which did not cause annoyance to other occupants of the building. The walls consisted of two layers of Celetex boarding separated by a layer of glass wool about 4 inches thick and the floor and ceiling were also padded with glass wool. The observation window consisted of two layers of glass about 2 inches apart. It was found that the reduction in overall level of the noise afforded by such a construction is approximately 20 db. The noise generator as issued to the Service consists of a scanning unit, power amplifier and loudspeaker. A close-up photograph of the scanning unit is shown in Fig. 2. A light beam is focused on an engine noise sound track on a glass disc mounted horizontally on the motor shaft. The modulated beam acts upon a photo-electric cell and the resulting voltage variations are amplified and transmitted to a large moving coil loudspeaker in the testing booth. The sound track was produced by recording on the ground the noise of a Lockheed Hudson bomber after transmission over a high-fidelity radio link. The original recording was made on a direct recording disc and a particularly uniform section was then re-recorded photographically on a sensitized glass disc in the form of a continuous track. Four types of noise, corresponding to different flying conditions (normal cruising speed, maximum boost, etc.) were recorded in this way. The output of the photo-electric cell goes to a straightforward resistance-capacity coupled valve amplifier capable of delivering 50 watts into an output circuit of 600 ohms impedance for an input voltage of 11 millivolts. The output of this amplifier is led through an impedance matching transformer to the energized Pamphonic moving coil speaker in the booth. This loudspeaker is capable of handling an input of 20 watts. The overall diameter of the stiff paper cone is 16 in. and the impedance of the moving coil 10 ohms. The complete noise channel is shown schematically in Fig. 3 and Fig. 4 shows a photograph of the interior of the booth. The sound track corresponding to normal cruising speed was selected 147 https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

Dickson, Simpson, Fry, Swindell, Brown as being most suitable for our purpose. Intensity and frequency characteristic measurements were made on the noise generated in the booth using the General Radio sound level meter and the S. T. & C. SCHEMATIC DIAGRAM OF COMPLETE ENqiNE NOISE CHANNEL. VALVE NOISE UNIT SCANNING UNIT CORRECTION UNIT MATCHINq TRANSFORMER. MAIN AMPLIFIER UNIT. 5PE1AKE.R. VOLTMETER. FIG. 3. noise analyser described in section 4d. The two microphones were mounted side by side on the axis of the speaker and about 18 in. from the centre of the cone. The gain of the noise amplifier was then increased as much as possible without producing undue napping of the speaker diaphragm. At this level the G.R. meter gave a mean reading of n o 148 https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

A NEW METHOD OF TESTING THE HEARING EFFICIENCY OF AVIATION CANDIDATES.— E. D. D. DICKSON, .). E. SIMPSON, ]). B. FRY, G. E. SWINDELL, R. E. C. BROWN. FIG. 4. Interior of testing booth. [face p. 148 https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

Testing Hearing Efficiency of Aviation Candidates phons. The intensity was not entirely uniform throughout a revolution of the glass disc but the difference between the extreme values was not more than 6 db. The intensity of the noise in half-octave bands was measured on the S. T. & C. noise analyser. The resulting spectrum is W O O S / S i i V M ,,.OI OJ. 3 A I ! V H 3 b 0-1-381030 t / oaft 7 /f 000'9 1 / / LOCKHEED HUOSON - SYNTHETIC. 1 CTRA,. ED HUOSON - IN FLlGt- »- UJ /) oooi OM'S OMV / y o r i / / coo's 1 / ooo's i // /p 0001 ffl oot M g 7/ u m o z ; ou a- i 1/ 009 — S — Id z a u o »»! Z B / a. u / { / \ Z 002 001 06 ] 0* ; OL -l-. - , — 03 00 , 00 09 Cit S o o 8 3 o i o o 'H3 O S / S i l V M 9 -OI 0 1 i n o 3Al.W"ljfc . n o « n o . r t S"I38!33Q shown in Fig. 5. The spectrum plotted inside the cabin of a Hudson in flight, using the same meter, is shown for comparison. It can be seen that our recorded noise is very similar to the actual noise in the region 150-1200 c.p.s. but that above and below this band the intensity falls off considerably. In order to compensate for this falling off a frequency 149 https://doi.org/10.1017/S0022215100007830 Published online by Cambridge University Press

Dickson, Simpson, Fry, Swindell, Brown correction unit was designed for use with the amplifier. The response curves corresponding to the various settings of the control switches on this unit are plotted in Fig. 6. When the unit was incorporated in the amplifier it was found that bass lift correction could not be applied without S a s S 000 01 oooc OC08 \ COOi \ ooo? I If & / oocs / if \ g I h M I \ a s \ i ft \ \\ \\ \ \ ootv ooot / \\ 7 / 0002 / / OOOI / V ooe ooe , v/r \ H\u A \ / / OCi 009 oos \ / / / / / / / 1/ // 1/ oov \

3. Articulation tests as the basis of a new test 142 Types of test, Sentence articulation tests, Word articulation tests, Syllable articulation tests, Sound articulation tests 4. Equipment for experimental tests 146 Provision of the noise background, Recording channel, Replay channels, Replay channel for speech, Replay channel for