Apphed Acoustws 27 (1989) 89-102

Microcomputer-based Singing Ability Assessment and Development

D a v i d M H o w a r d

Phonettcs and LmgmstJcs Department, Umverslty College London, Wolfson House, 4 Stephenson Way, London NW1 2HE, UK

&

Graham F Welch

Faculty of Education and Commumty Stu&es, Bristol Polytechmc, Bnstol BS6 6UZ, UK

(Recewed 3 August 1988, revised version recewed 21 November 1988, accepted 28 November 1988)

A B S T R A C T

A new mwrocomputer-based system ts descrtbed whtch has been developed for the assessment and development of smgmg abthty lt makes use o f a spectally developed hardware interface whwh esttmates the fundamental frequency o f a sung or spoken mput as the basts for (a) a measurement of vocal pttchmg accuracy (assessment), and ( b ) a vocal pttch &splay to provtde msual feedback (development) Results are dtscussed whwh are based on a study carrted out wtth the syatem in a Brtttsh prtmary school, and they mdtcate that this system ts effecttve m promotmg singing development

I N T R O D U C T I O N

There are many who take part in singing at all levels, f rom those who a r e

professional to those who simply stag for pleasure However, part icipatmn in singing actlwty ~s demed to many because their confidence or level o f singing abihty is undeveloped i For this group, any improvement in their

89 .4pphed ,4cousncs 0003-682X/89/$03 50 © 1989 Elsevier Soence Pubhshers Ltd, England Printed in Great Britain

90 David M Howard, Graham F Welch

voice pitching skills, lrrespectwe of starting point, could enable them to take some more active part m singing performance Research suggests that the necessary skills can be developed, provided that appropriate feedback is available, and the person IS prepared to put m appropriate practice

It is suggested by recent research 2-6 that singing ability can be characterized by certain stages along a continuum ofabdi ty This cont inuum ranges from those who find vocal pitch matching difficult, termed 'poor pitch' or 'uncertain' singers (e.g Refs 7-9) to those who have a multi-faceted singing ability with, for example, the ability to sing 'at sight' Singing development can be assisted with the use of appropriate visual feedback (e g Refs 10-14) This 'appropriate' feedback should be in a form which is readily accessible to the developing singer, who must be able to make proper sense and use of the information presented This is much easier m a private singing lesson, where individual attention is guaranteed, rather than in the school classroom, where there could be 30 children and perhaps a teacher with no background in voice training

The S INGAD ( S I N G i n g Assessment and Development) system 15 enables non-specialist teachers to develop singing skills in children and adults f rom the age of five upwards It can also be used successfully without any adult/ outside interaction The system makes use of pitch displays which have been designed to provide visual feedback in definite stages, thus allowing different levels of operation to match the varied voice pitching skills of users S INGAD is intended, in the first instance, to be used in primary and secondary schools The system makes use, at present, of the Acorn BBC microcomputer, as this is the machine which is ubiquitous in British primary and secondary schools This computer is used in conjunction with a specially developed hardware interface unit which processes a sung or spoken input, via ItS internal microphone, to provide an input to the microcomputer of voice fundamental frequency

THE S I N G A D SYSTEM H A R D W A R E

In its present configuration, the S I N G A D system reqmres a BBC microcomputer (model B or Master), with &sk drive, monitor and printer, as well as the S I N G A D hardware interface There have been three mare influences in the choice of this microcomputer, firstly, it is the mactune most commonly found m Brmsh primary and secondary schools at present, secondly, this machine provides a number of ports from which to choose for the interfacing of external devices and, thirdly, it incorporates a sound output system which is adequate for the provision of musical notes for use m both singing assessment and development

Staging abthty assessment and development 91

The SINGAD hardware interface is designed to estimate the fundamental frequency (F0) of a sung or spoken input The interface is small (currently packaged in a 110 x 45 x 70mm box) and battery-powered It makes use of an electret microphone which is currently mounted in the box, and there is a user 'gain' control and an output hght emitting diode (LED) to indicate the presence of a sound which is 'pitched' (a sung note or a voiced speech sound)

This interface has been developed as a specially modified version of the input stage 16 of hearing prostheses These are being used with the totally deaf 13 and profoundly deaf a by the EPI (external pattern input) group based in London and Cambridge Here, pitch reformation is provided as an adjunct to hp-readlng, and it is essential that the necessary processing occurs m real time so that the visual and auditory cues are available simultaneously It is also important that the F0 estimation device gives an appropriate output for both 'rough' and 'smooth' voice inputs, since the speech production of the deaf is often characterized by the increased use of rough voice It has been shown 19 that users can improve their own speech production skills if they can successfully monitor voice quality Thus the F0 estimation device operates m real time with no output smoothing

The SINGAD hardware interface estimates the F0 of a sung or spoken input by peak-picking 16 This involves locating the major peaks in the acoustic pressure waveform during phonatlon During sung pitches or voiced speech, the vocal folds vibrate and each vocal fold closure causes an acoustic lmpulse-hke excitation to the vocal tract Each major peak in the output acoustic pressure waveform represents one vocal fold closure, and the number of closures per second gives the FO in Hz

Peak-picking is just one of many methods which have been used over many years for F0 estlmauon and it has two special attractions in this application Firstly, it operates in real time, and visual feedback is most effective when provided synchronously with the acoustic output Secondly, no output smoothing is incorporated, and it is our experience that vocal pitches produced by young or inexperienced singers, or those whose singing confidence is yet to be gained, will often tend to have a broken-up or rough voice quality

The lnterconnecnon between the peak-picking device and the BBC microcomputer is made via the 'user port', and each peak-picking device output pulse generates an interrupt which ~s used as the basis for a fundamental period (Tx) measurement to an accuracy of 1 microsecond The reciprocal of this Tx value gives the cycle-by-cycle F0 value These measured Tx values are the input to either of the two software phases of SINGAD operation (assessment and development)

During the assessment stage and in one of the development stage options microcomputer sound output is required At present, in order to keep costs

92 David M Howard, Graham F Welch

Ftg. 1.

"0

(a)

i

I1,. ' ( b )

1. A. III 500Hz/d'v

Middle C (266 25 Hz) spectra (from Ref 25) (a) BBC microcomputer output alone, (b) BBC output with filtered output from momtor loudspeaker

to a minimum, use 1s made of the internal sound facdlty which ts standard with the BBC machme This is a four-channel system m which three of the channels give a penodtc output, and the fourth has an output which is based on norse It ts not possible to alter the spectral characteristics o f the sound output, but the F~J, mtenstty and duration are under user control

A measured output spectrum 2° for middle C (256 Hz) ts shown m Fig l(a), and It can be seen that It consists primarily of the odd harmomcs, with a modest additional contribution from the second and twelfth harmomcs The higher harmomcs dommate the spectrum, with the fundamental having a relatively low amphtude This output has a dtstmct 'buzzy' quahty, and tt has occasionally been found, during mformal trials, that users make octave errors when at temptmg to make a vocal match to the microcomputer- generated tones However, the momtor used has an audio channel avadable, and this has enabled use to be made of a simple passive filter stage (one-pole low-pass, cut-off frequency l k H z ) between the audio out from a BBC Master series m~crocomputer and the momtor The spectrum shown m Fig l(b) shows the acoustic spectrum, again for middle C, when the filter stage is used The result is a reduction m pitch ambiguity Two spectral effects can be

Smgmg abzhty assessment and development 93

noted the relative amphtude of the higher harmomcs has fallen with respect to the fundamental, and there is the addmon of the fourth and sixth harmomcs as distortion products

The reduction m pitch amblgmty can be explained m terms of the place theory of pitch perception 21 A frequency analysis of the input complex sound resolves the lower harmomcs up to about the seventh, and an analysis of the spacing between these is used as the basis for pitch perception An mvesUgatlon of the spacings between the harmomcs up to the seventh m the upper spectrum of Fig 1 shows that there is a basis for a pitch doubhng amblgmty, whereas m the lower spectrum, with the addmon of the fourth and sixth harmomcs, there is no such basis The spectra m Fig 1 were obtained by means of a microphone placed at the approximate position of the user's left ear, and it should be noted that the spectral charactenstlcs obtained m any practical situation when the filter is m use wdl be a function of the relative volume levels from the computer's internal loudspeaker and the momtor's loudspeakers, as well as the hstenmg posmon

The relative increase of the amphtude of the fundamental component improves the timbre of the output sound, and the addmonal distortion products lessen the ambiguity m the percewed pitch Amphtude chppmg would be another method for introducing even harmomcs, but the new timbre would tend to be more buzzy than before. A more costly solution would be to interface a synthesizer using the musical instrument digital interface (MIDI), then any sound which the synthesizer can produce could be used

THE SINGAD SYSTEM SOFTWARE

The SINGAD system consists of two phases, staging assessment and singing development. The menu and screen display software is written m BBC BASIC, and the 6502 assembler has been used to access the hardware interface and carry out the fundamental penod measurements The system is enUrely menu-driven, since it is pnmardy designed for users who are computer-naive, and every screen display incorporates the available options A block diagram of the SINGAD software is shown m Fig 2 Each mare menu (assessment or development) has an option to access the other menu and to revert to BASIC on completion The operation and purpose of each of the two phases is described below

Assessment phase

In the assessment phase, the BBC microcomputer generates up to 10 randomly ordered tones from either the 'treble' or 'bass' range sound sets

94 Dawd M Howard, Graham F Welch

ASSESSNEHT MENU DEUELOPMENT MENU

1. Run sesston 1, Run session 2, filter test parameters 2, Change soreen set 3, Results and d;sk 1/o 3, 81ter pltoh range 4, Ex;t to development 4, Ex;t to assessment 5, Catalogue d;;k 5, Extt to BNGIC 6, Exit to BRSZC

[]=J • ) I se22,"gl 1.2,":.'. I I ? I i / I ~enu I I se t tangl l /

t : l I Fig. 2. SINGAD software diagram (from Ref 25)

shown in Fig 3. After each note has been played, the user's sung response Is measured and tabulated (see F]g 4) in Hertz, octaves and semitones, and an absolute mean octave and semitone value is calculated by way of a summary For each sung response, the Hertz measurement is the mean of the first 255 T x measurements made wluch are considered to be 'vahd', i e each value falls within the allowed F0 range (30-1000 Hz) The results data can be printed and saved to &sk for further stahstlcal measurements to be made lfreqmred When stored, the random order, which vanes each time the program is run, is maintained to allow any ordering effects to be evaluated tf appropriate

The treble assessment range (see Fig 3) consists of the ten 'white' notes on a keyboard between A3 (220 Hz) and C5 (512 Hz), and the bass assessment range uses the ten wh]te notes between B2 (123 Hz) and D4 (293 Hz) The notes in use m the treble range have been chosen to include the p]tches

Fig. 3.

| 1 . , - , I I t I _ . I',~'~ J I I I J ~' - , , ,.

i - - : l I, I t A J I " I I I I I I I I - - l o _ ~ I~" I ' I I

Computer-generated tones used m SINGAD assessment (a) Tones used m the treble range, (b) tones used m the bass range

Staging a&hty assessment and development 95

SINGAD ASSESSMENT RESULTS TABLE *******************************

(C) DMH & CB (Vers: 2) Bass range; I0 notes; Len: Is; VoI:

DMH - May 1 9 8 8 at UCL

1 2 / 1 5 .

T e s t ( H z ) Sung ( H z ) O c t d i f S / t o n e d l f 178 000 1 7 8 . 6 3 5 0 . 0 0 5 0 . 0 6 2 2 6 6 , 3 0 0 2 6 9 . 4 6 9 0 . 0 1 7 0 . 2 0 5 1 2 5 . 7 0 0 1 2 4 . 0 3 9 - 0 . 0 1 9 - 0 . 2 3 0 1 3 3 . 3 0 0 1 3 2 . 3 6 3 - O . O l O - 0 . 1 2 2 1 4 9 . 7 0 0 1 4 8 . 8 9 8 - 0 . 0 0 8 - 0 . 0 9 3 2 5 1 . 3 0 0 2 5 2 . 4 6 1 0 . 0 0 7 0 . 0 8 0 1 9 9 . 6 0 0 1 9 8 . 8 0 7 - 0 . 0 0 6 - 0 . 0 6 9 1 6 7 . 7 0 0 1 6 6 . 5 2 8 - 0 . 0 1 0 - 0 . 1 2 1 2 9 9 . 9 0 0 3 0 2 . 2 9 7 0 . 0 1 1 0 . 1 3 8 2 2 4 . 3 0 0 2 2 5 . 5 3 0 0 . 0 0 8 0 . 0 9 5

ABSOLUTE MEAN DIFS" O.O10 O.121

Fig. 4. Example of SINGAD assessment ~sultstable

typically produced by poor-pitched child singers 22 These have been found to lie in the fifth between A3 (220 Hz) to E4 (330 Hz), thus the treble range starts at A3 (220 Hz) and extends upwards beyond this fifth to include notes frequently used in singing activities with children The notes in the bass range were intended to be an octave below those in the treble range, but the BBC sound generator range prevented this by one semitone, as seen in Fig. 3 When a user requests less than ten notes, these are delimited downwards thus a request for five notes would result In the five 'white' notes from A3 (220 Hz) to E4 (330 Hz) reclusive being used The volume and duration of the computer-generated tones can be altered While the spectrum of the output sound cannot be altered in this particular microcomputer, the amphtude characteristic can, and this is contoured sufficiently to avoid clicks and to give gentle attack and decay characteristics

The notes that the BBC microcomputer is capable of generating are defined to be quarter tones, tuned to the equal tempered scale based on A4 (440Hz) The fundamental frequency values which appear in the SINGAD results table as 'Test (Hz)' are taken from Waugh 2a These values have been confirmed by experiment as follows. The fundamental frequencies of each of the ten notes presented In the treble range SINGAD assessment (see Fig 3) were measured using a laboratory spectrum analyser 20 The results from this analysls are shown in Fig 5, where they are presented as a SINGAD assessment results table, the 'sung' values here refer to the fundamental frequency values measured by spectrum analysis The measurement accuracy for the spectrum analysis is shown at the foot of Fig 5, and the absolute mean semitone difference for the ten notes is less than

96 DavM M Howard, Graham F Welch

Test (Hz )

2 2 4 . 3 0 0 4 4 9 . 3 0 0 399 .800 3 3 5 . 8 0 0 266 .300 355 ,900 5 3 3 . 9 0 0 5 0 3 . 7 0 0 2 5 1 . 3 0 0 2 9 9 . 9 0 0

Sun9 (Hz)

2 2 4 . 5 0 0 4 4 9 . 5 0 0 3 9 9 . 0 0 0 3 3 6 . 2 5 0 2 6 6 . 2 5 0 3 5 6 . 2 5 0 5 3 4 . 0 0 0 5 0 5 . 0 0 0 2 5 1 . 2 5 0 3 0 0 . 0 0 0

ABSOLUTE MEAN SEMITONE DIFFERENCE:

S/tone d i f f .

0 . 0 1 5 O. 008

- 0 . 0 3 5 0 . 0 2 3

- 0 . 003 0 . 0 1 7 O. 003 0 . 0 4 5

- 0 . 0 0 3 O. 006

0 . O 1 6

ERROR IN SPECTRUM ANALYSIS MEASURE: )500Hz pluslmlnus 0.625Hz (500Hz p l u s / m i n u s 1 .250Hz

Fig. 5. Measurement of BBC-generated tones Column 1, stored refence values (TEST), Column 2, BBC output values measured by spectrum analysis (SUNG)

2% The repeatabihty between BBC m~crocomputers was Informally checked by measuring the FÙ of middle C as played by six different machines, and m every case the measured note was 266 25 Hz.

The SINGAD assessment phase is usually used at the start of a single session, or a number of sessions. In this way, progress with the use of the SINGAD development phase can be monitored

Development phase

The development phase is designed to enable the user to develop p~tchmg skdls via appropriate visual feedback Th~s feedback consists of a dotted 'pitch line' which crosses the screen from left to nght m four seconds, as a plot of the logarithm of F0 (calculated as the reciprocal of the measured T,, values from the peak-picking device) The sweep of the pitch line is mmated by pressing the space bar During non-sung or voiceless speech sounds the pitch hne tracks along the bottom of the screen (Fig 7) Subsequent pitch hnes can, if reqmred, be plotted without erasing the previous hnes using an overlay facihty

For vocal pitch to come under conscious control, the user must be able to make sense of the required task s Visual feedback can be used to provide an external correlate of vocal pitch movement, and can also act as a 'standard' for self-assessment Expenence with visual feedback encourages the development of an internahzed criterion of performance The SINGAD development phase provides visual feedback of various kinds which can be

Smgzng abdtty assessment and development 97

M r _ I

I , - v

t3 d "~ t)

~ l r e I 1 1 I W ~dlP I e I

1 2 3 4 5 6 7 8 Fig. 6. S INGAD screen ranges available m the development phase

matched to the user's current staging abdity (i e their part]cular level on the developmental continuum) Task &fficulty zs varied by selecting the number of visual targets to be &splayed over which the pitch line is superimposed

The first step In the development phase involves setting the F0 range on the momtor, and there are oght preset ranges available (see Fig 6) The mus]cal notes shown in Fig 6 define the upper and lower bounds of the &splay Settings 1 to 3 are in the 'treble' range, settings 4 to 6 are m the 'bass' range, and settings 7 and 8 fix the screen hmits to match those of the treble and bass clefs

The next step zs to select the level at which development work will be carried out A blank screen ]s probably most appropnate for those with least vocal p]tch development. A young chdd, for example, would usually begin wzth screen range setting 2, as this covers the pitch range within which poor- pitch singers are hkely to achieve most success at vocal pitch control Work w]th this screen could revolve speaking, singing or making any sound in order to observe the resulting p]tch line Sounds from everyday life could then be lm]tated, for example cows, sheep, cars or a fire engine (see Fig 7), the zdea being to develop an understanding that vocal pztch can be controlled and that cho]ces can be made Any vocal p]tch changes will be shown as a

• . . . - - .% -, •

SPACE=go; O ( / ] a p ) ; O ( u i t ) Fig. 7. S INGAD development blank screen wzth 'fire engine' pitch trace

98 Dawd M Howard, Graham F Welch

a. - P

) • . ° %

%

%

- .-.. 7 . ~P~,-..

"..._

J ( u g g l e ) ; SPACE=9o; O ( / l a p ) ~ Q ( u i t ) Fig 8. SINGAD development screen with three target pictures and 'successful' pitch trace

vertical movement m the pitch hne Conscious pitch control can be encouraged by attempting to 'draw' shapes on the screen, such as mountains or valleys, or an 'X' could be produced by overlaying a second pitch hne After th~s 'free play' stage w~th a blank screen, further control ~s encouraged by d~splaymg visual targets, the a~m being to 'hit' the target w~th the vocal p~tch hne

There are three libraries of pictures at present, and one, two, three or four p~ctures can be presented on the screen to act as targets These pictures can be juggled to alter their posmons on the screen, but if several are displayed together they will always be at &fferent screen heights, thus some change m pitch is reqmred to hit all the pictures m a multi-target display. Figure 8 shows a display of three pictures with an overlaid pitch hne which has successfully 'hit' all three targets The next option consists of a single picture which has an assocmted computer-generated tone that sounds when the space bar (or 'O' for pitch hne overlap) Is pressed prior to the pitch hne sweep begmnmg This provides the user w~th both an aural and a visual target, 1 e there is concurrent a b~-modal pitch model w~th wsual feedback Accurate vocal p~tch reproduction will result m the p~tch hne 'h~ttmg' the target (see F~g 9)

Further displays are available which use either a treble or a bass stave These allow interval pitching, wbrato, note onset and offset, and pitching m non-Western t radmons The poss~bflmes m th~s area remain relatwely untouched, since work with the S I N G A D system has, to date, concentrated upon staging assessment and development with children

Smgmg abthty assessment and development 99

J(uggIe); SPACE=9ot, O ( / ] a p ) ; Q(uit) Fig. 9. S INGAD 'b~-modal' development screen the computer generates a tone at the p~tch

reqmred to 'h~t' the rocket target

DISCUSSION

The SINGAD system has been the subject of a pilot study m London schools, 15 which suggested that the system could be effectzve m the development of staging ablhty m children As part of this study, a protocol was developed and the software was modified to essentially zts present form.

More recently, a larger scale long-term study has been completed 24 with a Bristol primary school class of seven-year-olds. This study was designed to mvesugate the effectzveness of the SINGAD system m promotmg smgmg development Inmally, the staging ablhty of the whole class was assessed usmg the SINGAD assessment software The 32 chddren were spht into three groups, with the vocal pitch accuracy m each being equally matched statistically. One group acted as a control and they followed a more tra&tlonal music program, smgmg songs as a group w~th gmtar accompamment The other two groups had regular SINGAD development sessions, one with and the other without adult mteracuon Adult mteractmn took the form of &scussmg the vzsual feedback and switching to other development screens as appropriate Each staging development session lasted ten minutes and there were seven sessions across the term The vocal pitch accuracy of the whole class was subjected to a second SINGAD assessment after one school term The control group showed a non- slgmficant ~mprovement m vocal pitching ablhty, whereas the other two (SINGAD) groups, one w~th and one without adult interaction, &splayed a slgmficant improvement. The vocal pitching ability Improvement m the group with adult interaction was shghtly better than that of the group who worked wzth the system without adult help, but the difference here was non- slgmficant This last finding is particularly interesting m that it suggests that

100 Davzd M Howard, Graham F Welch

the SINGAD system could have a posmve effect on staging development m a classroom without being dependent on vocal skdl development input from the teacher

In its present form, the SINGAD system is designed to work with pitching on the equal-tempered scale Although it has a relatwely inexpensive sound output, the BBC microcomputer could be programmed to produce p~tches at quarter semitone intervals for the assessment phase, and to the nearest plxel m the development phase This would allow the system to be used for work w~th, for example, tuning systems of other cultures A further extension to the system would be if the required sounds were produced by an externally connected MIDI (Musical Instrument Digital Interface) eqmpped synthesizer In thxs way any pitch could be produced, and the spectrum of the sound output could be adjusted as reqmred to enhance hstenmg interest, and to avoid the occasional octave pitching errors such as those described above The pitches of stlmuh could be more accurately defined and, m the development phase, bl-modal targets consisting, for example, of ghssan& which map onto a smoothly changing contour of target pitch rather than a single picture could be used

Other aspects of the staging voice are bee#nmng to be quantified (see e g, Ref 25), and it should be possible to provide the hardware interfaces for their apphcat~on in later generations of such visual feedback systems It should be stressed that the use of such systems is intended to free the professional voice trainer from the more routine, now quanufiable, aspects of staging voice production, and to enable the teacher to devote more time to the essentmlly unquanUfiable skdls such as performance and interpretation At the same t~me the SINGAD system enhances staging development m the class of the non-speclahst In either case, the students can gain basic voice skills m m~crocomputer-supported practice sessions

CONCLUSIONS

The SINGAD system has been developed for the assessment and development of staging ablhty It currently operates with a standard BBC microcomputer, and the system consists of a specially developed hardware interface which derives fundamental frequency from a speech or sung mput w~th a state of menu-driven programs SINGAD has been used with children in primary schools m Britain, and it has been shown that it can make a statistically s~gmficant contribution to the development of vocal p~tchmg skdls over a comparaUvely short time The system could also be used for advanced staging development, including work on pitching m non-Western t radmons

Smgmg abthty assessment and development 101

It ~s intended that future developments wdl include porting of the system onto other microcomputers, and making use of new graphics and sound- generation faohtles The use of MIDI-eqmpped synthesizers ~s being considered for the generation of sound targets, since they offer the possibility of continuous pitch changes which would serve as more complex auditory targets, as well as the production of different sound colours to combat octave pitching errors They could provide greater auditory mterest The basic S I N G A D system also offers itself for development as a system for music composition based on sung input, where the tuning system could be chosen to be equated exactly with the sung fundamental frequency, or fixed to the nearest note of whatever tuning system has been chosen. The mterface between human and machme is rapidly mowng towards being based upon our own prmopal commumcat~on channel, and the use of a voice input is the first fundamental step towards a machme w~thout a keyboard or a screen

A C K N O W L E D G E M E N T S

The authors would hke to thank Chris Rush and Rachel Gibbon for their work with the S I N G A D system m schools, Cohn Bootie for his machine code programming, Stuart Rosen for helpful d~scusslon, and all the chddren and staff from the schools m London and Bristol who helped to make the trials possible

R E F E R E N C E S

1 Mawhmney, T H & Cuddy, L L, A factor analytic investigation of tone deafness Paper presented to Music Education National Conference, Chicago, 1984

2 Gardner, J, Davldson, L & McKernon, P, The acquisition of song a developmental approach Documentary report to the Ann Arbor Symposium, Reston, Vlrglma MENC, 1981, pp 301-15

3 Dowhng, W J, Development of musical schemata in children's spontaneous singing In Cogmttve Processes m the Percepnon of Art, eds W R Crozier & A J Chapman Elsevier, Amsterdam, 1984

4 Welch, G F, Chddren's staging a developmental continuum of ability J Res Staging, 9(2) (1986), 49-56

5 Welch, G F, A developmental view of children's staging Br J Mustc Educ, 3(3) (1986), 295-303

6 Rles, N, An analysm of the characteristics of lnfant-chdd singing expressions replication report Can J Res Mustc Educ, 29(1) (1987), 5-20

7 Welch, G F, Poor pitch singing a review of the literature Psychol Muszc, 7(1) (1979), 50-8

8 Welch, G F, A schema theory of how children learn to stag In-tune Psychol Music, 13(1) (1985), 3-18

102 Dawd M Howard, Graham F Welch

9 Phlhps, K, Chdd voice training research J Res Staging, 8(1) (1984), 11-26 10 Galser, P E, A study of tone-matching techmques as remedml mstrucuon for

non-stagers PhD thesis, Oregon, 1961 11 Romaine, W B, Developing stagers from non-stagers--an lnvesUgation of a

speech control remedml procedure DEd thesis, Columbia Umvers~ty, New York, 1961

12 Cobes, C J , The condlhomng of a pitch response using uncertain singers Bull Councd Res Mustc Educ, 30 (1972), 28-9

13 Jones, M, Using a vertical-keyboard mstrument with the uncertain stager J Res Musw Educ, 27 (1979), 173-84

14 Welch, G F, Variability of prachce and knowledge of results as factors m learning to stag m-tune Bull Counctl Mustc Educ, 85 (1985), 238--47

15 Howard, D M, Welch, G F , Gibbon, R R & Bootie, C, The assessment and development of staging abd~ty--mmal results with a new system Proc Inst Acoust, 9(3) (1987), 159-66

16 Howard, D M & Fourcm, A J, Instantaneous voice period measurement for cochlear stimulation Electron Lett, 19 (1983), 76-8

17 Fourcm, A J, Douek, E E, Moore, B C J, Rosen, S M, Walllker, J R, Howard, D M, Abberton, E R M & Frampton, S, Speech perception with promontory stimulation Ann NY Acad Sct, 405 (1983) 442-8

18 Rosen, S M & Fourcm, A J, When less is more--further work Speech, Hearmg and Language Work m Progress Umverslty College London, 1 (1983), 1-27

19 Abberton, E R M., Fourcln, A J , Rosen, S, Walhker, J R., Howard, D M, Moore, B C J, Douek, E E & Frampton, S, Speech perceptual and productwe rehabilitation m electro-cochlear stimulation In Cochlear Implants, eds R A Schmdler & M M Merzemch Raven Press, New York, 1985,pp 527-37

20 Howard, D M & Welch, G F, A new microcomputer system for the assessment and development of smgmg abdlty Paper presented at Inst of Acoustics Conf on New Developments in Instrumentation and Computing m Acoustics, London, May, 1987

21 Moore, B C J, An Introductton to the Psychology of Hearing, 2nd edn Academic Press, London, 1982

22 Welch, G F, Vocal range and poor pitch staging Psychol Mustc, 7(2) (1979), 13-31

23 Waugh, I, Making Music on the BBC Computer Sunshine Books, London, 1984

24 Welch, G F, Rush, C & Howard, D M, The SINGAD (SINGIng Assessment and Development) system first apphcatlons m the classroom Proc Inst Acoust, 10(2) (1988), 179-85

25 Howard, D M & Lmdsey, G A, New iaryngograms of the staging voice In Proceedings of the 11th International Congress of Phonetws Sciences, Talhnn, USSR, 1987, Vol 5, pp 166-9