Class I. The letters T O S appear frequently both as initial letters and as final letters in their own words, with terminal O confined largely to short words. All three of these are very freely doubled.
Class II. The letters A I H appear frequently as initial letters, but far less frequently as finals, especially A I. Not one of these is readily doubled.
Class III. The letters E N R appear frequently as final letters, but far less frequently as initials. The letter E is very freely doubled; the other two not so often.
The following further observation might be made: When one of these letters changes its class, the least likely exchange is one occurring between classes II and III.
Now let us return to the foregoing cryptogram and consider the application of this information. A frequency count taken on this cryptogram will show that when its letters are rearranged according to their frequencies, they divide automatically into three rather clearly-defined groups, much like those of the normal frequency table. There are eight letters which outrank the rest, and these, named in the order of decreasing frequencies, are: R, D, V, S, F, Z, K, X. Presumably, then, most of these are substitutes for letters of the class E T A O N I R S H.
If an examination now be made of the terminal letters in the cryptogram, it will be found that, of the eight considered, the letters R D F have appeared at least once in both positions. These we may label class (a), as being good material for the originals t, o, s. It is found that the letters S Z have appeared at least once as initials, but not at all as finals. These we may label class (b), that is, good material for the originals a, i, h, except for a point which will be mentioned in a moment. The remaining three letters, V K X are found at least once as finals, but not at all as initials; these we will call class (c), good material for the originals e, n, r. Thus, we are enabled to begin our work by noting down the following possibilities:
| (a) R D F might represent (I) — T O S. | (Compare the facts: t, o, n). |
| (b) S Z might represent (II) — A I H. | (Compare the facts: a, c). |
| (c) V K X might represent (III) — E N R. | (Compare the facts: e, h, l). |
While such a classification is probably never 100% accurate, the writer has still to find a cryptogram (unless among the very badly manipulated “aristocrats”) in which at least part of the assumptions are not correct. We are dealing, however, with the very short cryptogram, in which a single occurrence of a letter in a given position can be regarded as of some importance.
Ordinarily, the most frequent letter of (c) will represent e, as it does here. This letter is famous as a final letter, and any printed page will show it at the end of 17 or 18 words in every hundred. There is not so clear a distinction between T and S of class I.
The most vulnerable of the groups, however, is (b). Of the three letters which may be represented here, two are vowels, concerning which we are to hear more, and not one is readily doubled. When Z, tentatively included in this group, is found to have been doubled near the beginning of a word, it is seen to be wrongly classified.