50th Anniversary of Establishment of Phonetics Dept.,
I have been asked to talk about the Technical and Experimental side of the Department's work from its birth to the early 1960's. John Laver and Bill Hardcastle will take up the story from about that time.
This is the very last opportunity we have, it seems to me, to look back and commemorate what was in many ways a unique department and I would be doing it, and the many people who served it, an injustice if I did not tie in a reasonable amount of background of facts and dates. But it has not been easy with my laconic diaries and lack of official records.
This is also perhaps the last time to pay proper tribute to a great man, a great Head of Department and a good friend, David Abercrombie.
So let us start at the beginning and open the curtains for the first part of our 50 year long play. We are looking into the 'rather drab basement' (as John Kelly described it) , in the very dark, very empty, bowels of Minto House in Chambers Street and it is 11 am on Tuesday the 5th of October 1948.
Two people sit, one each side of a small table. David Abercrombie, Lecturer in Phonetics, and myself. Both of us have our questions and answers and eventually dry up. Then after a little pause I am offered the job. I accept. Now I am become Technician of the Phonetics Department of the University of Edinburgh. I did not realise, then, of course, how very different my life was going to be. And nobody warned me then that I would have to give this first paper today 50 years and 20 days on.
One of my first tasks was to go down to England to the disc recorder company to learn how to make gramophone records, or discs as we called them, then to SOAS and UC to learn how to make artificial palates for Palatography and how to generate the right kind of smoke for the Kymograph. All of this was hardly Advanced Electrical Engineering, of course, and having been involved in real State of the Art communications in the Middle East Phonetics technology seemed more than a little old-fashioned.
By the way, I started off as Anthony and quite soon became Tony. And in no time at all, Mr. Abercrombie became David. That was his kind of department - which was more than a little unusual in those days.
So I was Tony by the time Betsy [Mrs. Elizabeth T. Uldall] blew in a little later from goodness knows where to brighten up our drab basement like a ray of American sunshine. She lit it up with her great big smile and showed us that in professional hands the kymograph was, indeed, a scientific instrument that could tell us a great deal -- black faces or no.
Now sometime after when we had installed everything that we thought we could possibly need, I remember asking David, when, when, when, are we really going to start? When do we get students? But he only said 'Patience -- Patience' with his small smile and before long we did have students -- hundreds of them I think, everywhere one turned there was a student, each with their own unique and very important question. It seems fairly certain that our first teaching session was 1949-50. Can anyone remember?
Friday, I seem to remember was the main Lab Day. How many of you here today can look back to those days, when we had at least ten experiments going on at the same time in the big lab with the high windows?
The staff of the department of that time must, I think, look with a quiet pride at the range of teaching and practical lab work that was offered and I would like to think that our department became the model for many other departments of phonetics throughout the world.
So having set the scene we can now consider some of the technical advances made by the department then.
First of all, in Palatography. As we inherited it, this involved taking a wax impression of the palate of each subject and then, from that, making a black artificial plastic palate. To make a palatogram the investigator dusted this artificial palate with french chalk, put it into his mouth and said what he was interested in. He then took it and if the wipe-off was satisfactory he photographed it as a permanent record. Somewhat time-consuming, considering the number of students we had.
We simply had to simplify the process. The subject now sprayed his own palate with a mixture of charcoal and chocolate. He then said what he had to say, checked the wipeoff and photographed it. Just like that.
Freddie MacAuley will remember that in just one day we processed, I think, 808 palatograms for a Gaelic paper of his. Did it ever get published?
Disc recording was the only means of recording that we had then, of course, and the repeating of a sample required the lifting of the pick-up and putting it back in the same groove as before. A slow process for only one repeat and also a rather distracting human mechanical process.
And when we changed to tape recording as standard, repeating a sample took even longer as one had to stop the machine, switch it to wind back, stop it again, switch to play and start it again for just one repeat. So we more or less had to invent the Tape Repeater -- and had to patent it or someone else would have. This was a simple auxiliary loop machine where what was being heard from the master tape was always being recorded and wiped off as the loop went round. When one stopped the main machine one automatically switched to the loop and it played its sample for ever and ever and ever and ever.
This may not appear a big advance nowadays when the computer will do anything you ask -- if you know what to ask -- but in those days the Tape Repeater was a tremendous help in the transcription of phonetic text and the folksong and folkstory that we had collected in the field.
That brings me to another of the demands on departmental time. Down in the basement of the Awful Ferguson Building in George's Square lie the recordings of the Linguistic Survey of Scotland and an unique collection of Scots and Gaelic Folksong and Story.
From our quite early days people from our department and many other departments with our help collected this material summer and winter in all corners of Scotland, in all kinds of weather and all kinds of conditions. With his funding for travel and equipment and his foresight and encouragement this collection is one of Professor Angus MacIntosh's greatest achievements but Scotland is being a little slow in acknowledging it, I am sorry to say.
Professor MacIntosh sends his Best Wishes.
This collecting began, for me at least, with the first rather experimental trip going with Miles Dillon to the far West, -- and I remember his delight in talking, in Irish of course, to the gaelic fishermen from the Islands in the Columba Hotel in Oban. There were many trips, too, with Derick Thomson (who became Professor of Celtic at Glasgow) collecting in the Highlands and Islands from Newtonmore to the Butt of Lewis, night and day -- but mostly night. And with Calum Maclean, brother of Sorley the poet, who was one of our great Gaelic Folkstory Collectors. I remember one dark snowy winters night in South Uist going to a Black House with him hoping to get a recording of a Duan, an ancient gaelic chant. This was the Duan -- Duan na Cearschact -- and with his unique approach Calum did get the local version from this 100 year old man, by candlelight, while the wind blew hard outside.
But there are many other collections down there, of course, in that basement -- Modern Scots (The Jack Aitken Collection), Bothy Ballads from the North East with Hamish Henderson. But not, I am sorry to say, the Horseman's Grip and Word), Scots Dialect with Hans Christian Uldall and Ian Catford -- collected from the East Neuk of Fife to Avoch -- and much else. I even ventured with John Wooley (Whatever happened to him?) to furthest Wigtonshire where they still (at least at that time) put an iron nail in the butterchurn to keep the fairies away -- or was it the Devil?
We even, Angus MacIntosh and I, collected Modern Folklore in the old department down in Minto House -- from Charles Laughton, (who was appearing in the Empire theatre round the corner at the time). He was very interested in our recordings which illustrated the style of delivery used by the old gaelic storytellers. And afterwards he made quite marvellous recordings, for us, of 'America is a place of trains' and 'Shedrach Meshach and Abednego. (Which you can hear if you go to see the video recording of Eye on Research in the basement of the Ferguson Building today).
Let me stress that this is a quite large and quite unique collection (the gaelic section of which I propose should be called the MacIntosh Collection) will be of enormous importance to our coming generations. But these tapes and disks must be deterioriating however good the conditions are down there. Financial and technical support is urgent and essential for re-recording as soon as possible on CDs. The Irish government have already tackled this problem and part of their collection is on sale to their general public.
But the function of a Phonetics Department is Phonetics first and foremost and within two years of our birth we jumped into the present and became the envy of all other departments with the acquisition of a Kay Electric Sound Spectrograph. The department was now launched on a new kind of scientific career and there was more of a balance between Traditional phonetics and a quantitative approach to speech research.
With this very powerful means of acoustic analysis, we were at last able to study the curious dance of human articulation, and follow the intricate patterns of the vocal tract's ever-changing movement.
Now let me diverge for a minute to talk about one of the very serious problems that the world inherited after the war; this was the almost complete lack of speech transmission channels. Before 1939 there was really very little spoken communication between nation and nation -- just a few international cables and wireless channels whose capacity was extremely limited. Few people had a telephone and there were only manual town switchboards.
Suddenly after the war everybody wanted to talk to each other. There was an enormous demand for communication and it had to be speech communication.
While, it is true, new cables were being laid they were enormously expensive and though wireless transmission techniques were being developed it was all very very slow. Even in the nineteen fifties an old second-hand manual telephone exchange was rebuilt in George Street here in Edinburgh.
Many attempts had been made to find ways of reducing the channel capacity required for the transmission of speech. With statistical methods of frequency, amplitude and time compression bandwidth compression ratios of about 5:1 can be obtained; while analysis-synthesis methods and systems based on the recognition of phonetic units were thought might achieve an even better compression ratio of between 10:1 and 100:1.
One possible answer was to take advantage of what had been learned about the speech signal itself. Information Theory seemed to show that the transmission rate of information in speech was in fact low. Why not then, instead of sending the whole speech signal, analyse it into parameters which describe it? In other words why not simply transmit instructions to an analogue of the human speaking mechanism at the far end that would say what had been said at the sending end.
This was the thinking of a very clever engineer, Walter Lawrence, of the Signals Research and Development Establishment in Christchurch, England. And in 1952 he invented the Parametric Artificial Talking machine, known more familiarly as PAT which was the first dynamic synthesiser to be controlled by a set of parameters each describing an essential factor of the human vocal tract. It demonstrated quite clearly that intelligible speech could be generated in this way and that the parameters need vary only slowly.
The basic information that would have to be sent would be (1) is the speech at this instant voiced and if so what is its pitch and amplitude?, (2) if it is voiced what are the Formant Frequencies of the Vocal Tract Transmission Function? and (3) if it is consonantal noise what is its amplitude?
An international colloquium was organised at SRDE (The Methods and purposes of speech synthesis. SRDE Report No. 1100 March 1950) by Lawrence in 1955 which David Abercrombie attended. He immediately realised the machine's potential for phonetic research and before long we were asked to take part in a joint research contract. I then went down to SRDE and built a brother (or was it a sister?) for the original Pat and set it up here.
But the evaluation of the performance of band-width compression systems does present many phonetic problems and the years following were a period of intense research effort with synthesis techniques being used to provide information about the perception of the various acoustical factors of speech. Many people were involved and here I should mention the achievements of Betsy who synthesised the whole of "The North Wind and the Sun" and made a wonderful film of the vocal cords.
The direction of this work was first under Peter Strevens and I remember how Peter and I took Pat, lock stock and barrel, to a Phonetics congress in Oslo and how their customs asked for an enormous amount of money in dues so that they could be sure we would take it away again. Frances Ingemann who came from the University of Kansas and Haskins New York was Pat's second master or mistress. We had visitors and research workers from all over the world and we returned many of these visits. Gunnar Fant from Stockholm, Ken Stevens from MIT and Haskins not forgetting Victor Jassem from Poznan Poland who wrote a very nice paper with John Morton of the MRC. Frances Ingemann even visited the Russian Lab in Leningrad when this was a very difficult thing to do.
Here I would like to pay tribute to Walter Lawrence as a great engineer and to Peter Strevens -- PAT's first Tutor, both of whom, sadly, are no longer with us.
But to get back to the original bandwidth problem of speech engineering: to perform any clever trick the first essential is to catch your rabbit, the rabbit in this case being these seemingly simple control signals that would have to be extracted in real time from real raw speech. Now, as we talk to each other, the human ear and brain performs this process of extraction of information rather well. Deriving these signals from raw speech by electronic methods was, however, found to be a very very difficult problem indeed. The sad fact is that, even today, speech recognisers do not appear to be able to deal with normal running speech.
But the department had its Moment of Glory when we were asked by the B.B.C. if we would like to be one of the scientific centres portrayed on their Eye on Research programme. We could hardly say no. However, when we received their draft script we were somewhat less than overjoyed with it. So we said that as we had a Government Ministry of Supply Contract the programme would have to be approved by that Ministry. All white lies, of course, but it solved the problem and I think everybody that was involved in the writing of the script and making the actual programme thought we did a good job. But you can judge that for yourselves because we have set up a viewing room in the Ferguson Building.
My life changed considerably a year or two after that when I went off to the University of California at Los Angeles with Wife and Family at the invitation of Peter Ladefoged who had accepted the new Chair of Phonetics there. And by the time Nan and I and the kids came back home to Scotland a year or so later I had a developed a new and very strong interest in the clinical and physiological aspects of speech.
At that time, luckily, research grants were relatively easy to get and once I had acquired a fibre-optic laryngoscope a Voice and Speech Disorders Clinic was set up in the Royal. We also developed a procedure for the clinical injection of teflon into the vocal cord under fibreoptic observation. By the way, it is ironic that I should have had to import such a laryngoscope into Scotland when it was James Logie Baird, our great Scots inventor, who invented fibreoptics. (James Logie Baird, Brit. Patent No. 20,969 1927. An Improved Method of and Means for Producing Optical Images)
The development of pressure transducers allowed great improvements in subglottal pressure measurement and new airflow transducers allowed more accurate measurement in clinical cleft-palate speech assessment.
I made a number of interesting and curious investigations in these years. I even made measurements on the vibrating cadaver larynx and for quite a number of years we had a very fine Clinical team in the Sick Kids assessing speech before and after surgery. I also did a little Stammering assessment. But I think I found my real niche and still continuing interest when I began to look at the respiratory aspects of speech.
I said the curtains for the Phonetics Department play opened in 1948: -- they closed again, at least for me, with David's death in 1992. Sometime before then - I do not think it could have long before -- I heard that David was in the Infirmary and I found him there, in a public ward full of noise and bustle. He managed a small smile, we chatted for a while and then I left. But I went away feeling very sad because he looked so wan and tired. I am sorry that I was not here at the time of David's funeral and that that visit should be my last memory of him
But maybe the history of our Phonetics Department is not so very different from that of many other great academic departments of Science. In David Abercrombie it had its creator and superb teacher and continued a proud tradition from Henry Sweet and Daniel Jones forward proudly. It had its many years of glory in the firmament of academia and then lost its independence. The measure of its success, however, can perhaps best be judged (not by me -- I have never claimed to be a Phonetician) but by those who benefited from its teaching. One would like to think that out there in the many continents of our world the influence of that teaching is still great and confirmed by the quality of the phoneticians David Abercrombie left behind to continue his work.
J. Anthony, C.Eng. M.I.E.E. Ph.D. formerly Sen Lecturer.
Note: The main underlying necessity for the work in synthesis and recognition of speech has passed, of course, but the research position reached now is summed up, I suggest, in these two references.
(1) 1981 The Cognitive Representation of Speech. ed T. Myers, J. Laver and J. Anderson. N. Holland Pub Co.