302 CRYPTOGRAPHS AND CIPHERS [CH. XIV 



especially if the discovery of the purport of the message is, 

 after a few hours, immaterial. 



The key to ciphers of this type may usually be found by 

 using tables of the normal frequency with which letters may be 

 expected to occur. Such tables, and other characteristic features 

 of the English, French, German, Italian, Dutch, Latin, and Greek 

 languages, were given by D. A. Conrad in 1742*. His results 

 have since been revised, and extended to Russian, Spanish, and 

 other tongues. In English the percentage scale of frequency of 

 the letters is approximately as follows: — e, 120; t, 9'4; a, 7 - 8 

 o, 7-5; i, 7-4; n, 73; s, 68; r, 59; h, 57; d, 3-9; I, 36; u, 30 

 c, 2-8; m, 27; /, 25; p, 1-9; g, 1*8; y, 18; b, VI; w, 1-7; v, 11 

 k, 06; j, 03; q, 03; x, 0'3; z, 02. The order of frequency for 

 combinations of two letters is th, he, in, an, on, re, ti, er, it, nt, 

 es, to, st; of three letters is the, ion, &c, &c; of four letters is 

 tion, that, &c, &c. ; and of double letters is tt, ss, &c, &c. Other 

 peculiarities, such as that h, I, m, n, v, and y, when at the begin- 

 ning of a word, must be followed by a vowel, that q must be 

 followed by u and another vowel, have been classified and are 

 important. I need not go here into further details. Unless, 

 however, the message runs to 400 words or more, we cannot 

 reasonably expect to find the scale of frequency the same as 

 in Conrad's Table. 



In ciphers of this class it is especially important to avoid 

 showing the division into words, for a long word may easily 

 betray the secret. For instance, if the decipherer has reason 

 to suspect that the message related to something connected 

 with Birmingham, and he found that a particular word of ten 

 letters had its second and fifth letters alike, as also its fourth 

 and tenth letters, he would naturally see how the key would 

 work if the word represented Birmingham, and on this hypo- 

 thesis would at once know the letters represented by eight 

 symbols. With reasonable luck this should suffice to enable 

 him to tell if the hypothesis was tenable. To avoid this risk it 



* Gentleman's Magazine, 1742, vol. xn, pp. 133—135, 185—186, 241—242, 

 473 475. See also the Collected Works of E. A. Poe in 4 volumes, vol. I, p. 30 



et sej. 



