synthesizer was a windbox driven by a bellows. One output of the windbox led to a reed, to simulate vocal-cord excitation; the reed was followed by a short neck and a bell-shaped, rubber mouth. Deforming the mouth changed the quality of the sound made by the reed. Projecting from the neck were two tubes which could be opened to make nasal sounds. Other outputs of the box included special passages for the fricatives [s] and . There was also a lever to modulate the vibration of the reed for trilled [r], and an auxiliary bellows for aspiration of voiceless stops. A human operator played the synthesizer like a musical instrument; he pumped the bellows with his right arm, operated the various levers and tubes with his right hand, and manipulated the rubber mouth with his left hand.

Von Kempelen claims to have synthesized a number of short utterances in various languages ('Leopoldus Secundus', 'vous êtes mon ami'). In 1923, Paget operated a copy of the synthesizer built by Wheatstone (1837) and was able to produce a few isolated words (Paget 1930: 19). But whatever the quality of the synthesis, one cannot fail to be impressed by the insights into the nature of speech production reflected in the design of the synthesizer and manifest in von Kempelen's monograph. He understood the basic relationship between the larynx and the supraglottal cavities, and realized the special problems posed by nasals and fricatives. He understood also the importance of what Fant et al. (1963) were later to call 'synthesis strategy': a set of techniques for producing the various classes of sounds which exploits the possibilities of a particular synthesizer and minimizes its limitations (thus von Kempelen made an [f] by closing all regular outlets from the windbox and building up enough pressure inside to force air through the leaks in the box!). The obvious limitations of his work were the need to use a mechanical system, the acoustic properties of which were neither easily predictable nor readily alterable; and the use of a human operator for dynamic control, with the consequence that the 'rules' were not explicit, but were rather part of the operator's art. Interestingly enough, the Abbé Mical, a contemporary of von Kempelen's, is supposed to have built a synthesizer controlled by a pinned cylinder, such as is used in a music box. Wheatstone (1837:40-1) considered and dismissed the possibility of fitting his copy of von Kempelen's synthesizer with a control device of this sort:

It would be a very easy matter to add either a keyboard or a pinned cylinder to De Kempelen's instrument, so as to make the syllables which it uttered follow, each with their proper accentuations and rests; but unless the articulations were themselves more perfect, it would not be worth the trouble and expense.

During the century after von Kempelen and Mical, other manually operated, mechanical speech synthesizers were developed. Besides Wheatstone's copy of von Kempelen's synthesizer there was, for instance, Faber's Euphonia (Gariel 1879),