Abstract :
A serious problem which occurs in the development of large or otherwise costly speech communication systems (the term includes, but is not confined to, public telephone systems) is the achievement of a consistent overall design that is economical and yet never requires users to exert unreasonable amounts of mental or vocal effort in conversation. The ideal thus simply expressed is notoriously difficult to achieve and the problem is of long standing in the history of telephony. The paper reconsiders this problem and sets out a coherent `design¿ procedure by which studies of rating (providing figures of merit in terms that are meaningful for engineering design purposes) and user evaluation (study of merits of speech links for particular applications) can be conducted to yield closely complementary information. These two aspects, `rating¿ and `user evaluation¿, are concerned respectively with individual speech links and with systems or networks viewed as aggregates of such links. A further type of study, also important, deals with estimating the grade of performance which a complete network, designed to a given transmission plan, will provide; this can be described in statistical terms such as the proportion of calls of various types whose transmission should yield various levels of satisfactoriness to users. This last type of study, conveniently known as `network transmission performance¿, is of special value in the long-term planning and operation of large systems.
